llama 7B
1431 records across 166 hardware.
| hardware | model | quant | backend | tok/s | pp tok/s | W | ttft s | ctx/tg | date | class | flags | source |
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| H200 | llama 7B | Q4_0 | llama.cpp (CUDA) | 337.6 | 11633.96 | – | – | tg128 | 2025-11-15 | community | – | source |
| H100 | llama 7B | Q4_0 | llama.cpp (CUDA) | 328.7 | 11564.03 | – | – | tg128 | 2025-11-15 | community | – | source |
| B200 | llama 7B | Q4_0 | llama.cpp (CUDA) | 315.85 | 11510.56 | – | – | tg128 | 2025-10-25 | community | – | source |
| GeForce RTX 5090 | llama 7B | Q4_0 | llama.cpp (CUDA) | 315.3 | 17107.92 | – | – | tg128 | 2025-12-04 | community | – | source |
| GeForce RTX 5090 | llama 7B | Q4_0 | llama.cpp (CUDA) | 314.0 | 14805.32 | – | – | tg128 | 2025-12-04 | community | – | source |
| GeForce RTX 5090 | llama 7B | Q4_0 | llama.cpp (CUDA) | 312.33 | 16070.88 | – | – | tg128 | 2025-11-08 | community | – | source |
| Radeon RX 7900 XTX | llama 7B | Q4_0 | llama.cpp (Vulkan) | 310.91 | 6508.13 | – | – | tg128 | 2026-03-18 | community | – | source |
| Radeon RX 7900 XTX | llama 7B | Q4_0 | llama.cpp (Vulkan) | 308.88 | 6695.94 | – | – | tg128 | 2026-03-18 | community | – | source |
| H200 | llama 7B | Q4_0 | llama.cpp (CUDA) | 307.79 | 10428.65 | – | – | tg128 | 2025-11-15 | community | – | source |
| H100 | llama 7B | Q4_0 | llama.cpp (CUDA) | 303.08 | 10162.9 | – | – | tg128 | 2025-11-15 | community | – | source |
| GeForce RTX 5090 | llama 7B | Q4_0 | llama.cpp (CUDA) | 301.92 | 15505.79 | – | – | tg128 | 2025-12-09 | community | – | source |
| GeForce RTX 5090 | llama 7B | Q4_0 | llama.cpp (CUDA) | 300.4 | 14970.15 | – | – | tg128 | 2025-10-23 | community | – | source |
| GeForce RTX 5090 | llama 7B | Q4_0 | llama.cpp (CUDA) | 299.92 | 14036.03 | – | – | tg128 | 2025-11-07 | community | – | source |
| GeForce RTX 5090 | llama 7B | Q4_0 | llama.cpp (CUDA) | 299.85 | 14013.74 | – | – | tg128 | 2025-11-07 | community | – | source |
| Radeon RX 7900 XTX | llama 7B | Q4_0 | llama.cpp (Vulkan) | 298.5 | 6695.4 | – | – | tg128 | 2026-03-20 | community | – | source |
| GeForce RTX 5090 | llama 7B | Q4_0 | llama.cpp (CUDA) | 298.18 | 13951.14 | – | – | tg128 | 2025-11-07 | community | – | source |
| GeForce RTX 5090 | llama 7B | Q4_0 | llama.cpp (CUDA) | 297.98 | 14131.49 | – | – | tg128 | 2025-12-04 | community | – | source |
| GeForce RTX 5090 | llama 7B | Q4_0 | llama.cpp (CUDA) | 297.87 | 14988.21 | – | – | tg128 | 2025-12-04 | community | – | source |
| B200 | llama 7B | Q4_0 | llama.cpp (CUDA) | 297.68 | 10531.52 | – | – | tg128 | 2025-10-25 | community | – | source |
| GeForce RTX 5090 | llama 7B | Q4_0 | llama.cpp (CUDA) | 294.95 | 13964.52 | – | – | tg128 | 2025-11-08 | community | – | source |
| GeForce RTX 5090 | llama 7B | Q4_0 | llama.cpp (CUDA) | 290.02 | 14073.41 | – | – | tg128 | 2025-10-23 | community | – | source |
| RTX Pro 6000 Blackwell | llama 7B | Q4_0 | llama.cpp (CUDA) | 289.3 | 14463.16 | – | – | tg128 | 2026-06-11 | community | – | source |
| GeForce RTX 5090 | llama 7B | Q4_0 | llama.cpp (CUDA) | 287.55 | 13451.73 | – | – | tg128 | 2025-12-09 | community | – | source |
| RTX Pro 6000 Blackwell | llama 7B | Q4_0 | llama.cpp (CUDA) | 287.29 | 13403.7 | – | – | tg128 | 2026-05-08 | community | – | source |
| GeForce RTX 5090 | llama 7B | Q4_0 | llama.cpp (CUDA) | 286.51 | 16195.14 | – | – | tg128 | 2025-09-26 | community | – | source |
| GeForce RTX 5090 | llama 7B | Q4_0 | llama.cpp (CUDA) | 285.51 | 15602.75 | – | – | tg128 | 2025-11-07 | community | – | source |
| GeForce RTX 5090 | llama 7B | Q4_0 | llama.cpp (CUDA) | 284.03 | 16053.65 | – | – | tg128 | 2025-11-07 | community | – | source |
| GeForce RTX 5090 | llama 7B | Q4_0 | llama.cpp (CUDA) | 283.69 | 15915.61 | – | – | tg128 | 2025-11-07 | community | – | source |
| RTX Pro 6000 Blackwell | llama 7B | Q4_0 | llama.cpp (CUDA) | 281.11 | 16618.98 | – | – | tg128 | 2025-09-05 | community | – | source |
| H100 | llama 7B | Q4_0 | llama.cpp (CUDA) | 280.74 | 11263.29 | – | – | tg128 | 2025-09-05 | community | – | source |
| GeForce RTX 5090 | llama 7B | Q4_0 | llama.cpp (CUDA) | 280.28 | 13143.45 | – | – | tg128 | 2025-11-07 | community | – | source |
| Radeon RX 7900 XTX | llama 7B | Q4_0 | llama.cpp (Vulkan) | 278.22 | 6287.12 | – | – | tg128 | 2026-03-20 | community | – | source |
| GeForce RTX 5090 | llama 7B | Q4_0 | llama.cpp (CUDA) | 277.21 | 14142.08 | – | – | tg128 | 2025-09-26 | community | – | source |
| RTX Pro 6000 Blackwell | llama 7B | Q4_0 | llama.cpp (CUDA) | 275.79 | 18012.64 | – | – | tg128 | 2026-04-02 | community | – | source |
| RTX Pro 6000 Blackwell | llama 7B | Q4_0 | llama.cpp (CUDA) | 275.76 | 16652.18 | – | – | tg128 | 2025-08-11 | community | – | source |
| RTX Pro 6000 Blackwell | llama 7B | Q4_0 | llama.cpp (CUDA) | 274.2 | 14854.63 | – | – | tg128 | 2025-09-05 | community | – | source |
| RTX 5090 | llama 7B | Q4_0 | llama.cpp (Vulkan) | 273.68 | 11796.38 | – | – | tg128 | 2025-10-11 | community | – | source |
| RTX Pro 6000 Blackwell | llama 7B | Q4_0 | llama.cpp (CUDA) | 271.26 | 12242.46 | – | – | tg128 | 2026-05-08 | community | – | source |
| RTX PRO 6000 | llama 7B | Q4_0 | llama.cpp (CUDA) | 270.0 | 11443.35 | – | – | tg128 | 2026-01-22 | community | – | source |
| RTX PRO 6000 | llama 7B | Q4_0 | llama.cpp (CUDA) | 269.96 | 12231.9 | – | – | tg128 | 2026-01-22 | community | – | source |
| RTX Pro 6000 Blackwell | llama 7B | Q4_0 | llama.cpp (CUDA) | 268.52 | 14401.14 | – | – | tg128 | 2025-08-11 | community | – | source |
| H100 | llama 7B | Q4_0 | llama.cpp (CUDA) | 267.81 | 9918.34 | – | – | tg128 | 2025-09-05 | community | – | source |
| GeForce RTX 5090 | llama 7B | Q4_0 | llama.cpp (CUDA) | 266.7 | 15094.35 | – | – | tg128 | 2025-11-07 | community | – | source |
| RTX Pro 6000 Blackwell | llama 7B | Q4_0 | llama.cpp (CUDA) | 264.66 | 15928.07 | – | – | tg128 | 2026-04-02 | community | – | source |
| RTX 5090 | llama 7B | Q4_0 | llama.cpp (Vulkan) | 263.63 | 10381.64 | – | – | tg128 | 2025-10-11 | community | – | source |
| RTX Pro 6000 Blackwell | llama 7B | Q4_0 | llama.cpp (CUDA) | 260.42 | 12742.48 | – | – | tg128 | 2026-06-11 | community | – | source |
| GeForce RTX 5090 | llama 7B | Q4_0 | llama.cpp (CUDA) | 248.57 | 16041.54 | – | – | tg128 | 2025-08-02 | community | – | source |
| GeForce RTX 5090 | llama 7B | Q4_0 | llama.cpp (CUDA) | 239.62 | 14751.98 | – | – | tg128 | 2025-08-02 | community | – | source |
| H100 | llama 7B | Q4_0 | llama.cpp (CUDA) | 229.85 | 8457.52 | – | – | tg128 | 2025-09-05 | community | contradiction | source |
| RTX Pro 5000 Blackwell | llama 7B | Q4_0 | llama.cpp (CUDA) | 226.31 | 11046.55 | – | – | tg128 | 2026-02-28 | community | – | source |
| H100 | llama 7B | Q4_0 | llama.cpp (CUDA) | 219.23 | 7350.98 | – | – | tg128 | 2025-09-05 | community | contradiction | source |
| A100 | llama 7B | Q4_0 | llama.cpp (CUDA) | 215.11 | 5422.2 | – | – | tg128 | 2025-11-15 | community | – | source |
| RTX Pro 5000 Blackwell | llama 7B | Q4_0 | llama.cpp (CUDA) | 213.54 | 9952.91 | – | – | tg128 | 2026-02-28 | community | – | source |
| A100 | llama 7B | Q4_0 | llama.cpp (CUDA) | 200.9 | 5285.96 | – | – | tg128 | 2025-09-05 | community | – | source |
| A100 | llama 7B | Q4_0 | llama.cpp (CUDA) | 200.05 | 4967.14 | – | – | tg128 | 2025-11-15 | community | – | source |
| A100 | llama 7B | Q4_0 | llama.cpp (CUDA) | 196.26 | 5295.94 | – | – | tg128 | 2025-09-05 | community | – | source |
| Radeon RX 7900 XTX | llama 7B | Q4_0 | llama.cpp (Vulkan) | 195.3 | 3332.9 | – | – | tg128 | 2025-11-05 | community | – | source |
| RTX 5080 | llama 7B | Q4_0 | llama.cpp (Vulkan) | 192.17 | 8054.59 | – | – | tg128 | 2026-01-31 | community | – | source |
| RTX 5080 | llama 7B | Q4_0 | llama.cpp (Vulkan) | 192.17 | 8054.59 | – | – | tg128 | 2026-01-31 | community | – | source |
| GeForce RTX 4090 | llama 7B | Q4_0 | llama.cpp (CUDA) | 191.57 | 12506.97 | – | – | tg128 | 2025-08-24 | community | – | source |
| Radeon RX 7900 XTX | llama 7B | Q4_0 | llama.cpp (Vulkan) | 191.28 | 3531.93 | – | – | tg128 | 2025-11-05 | community | – | source |
| Radeon RX 7900 XTX | llama 7B | Q4_0 | llama.cpp (Vulkan) | 190.92 | 3889.54 | – | – | tg128 | 2026-09-10 | community | – | source |
| A100 | llama 7B | Q4_0 | llama.cpp (CUDA) | 190.88 | 4849.53 | – | – | tg128 | 2025-09-05 | community | – | source |
| RTX 5080 | llama 7B | Q4_0 | llama.cpp (Vulkan) | 190.78 | 7022.52 | – | – | tg128 | 2026-01-24 | community | – | source |
| RTX 5080 | llama 7B | Q4_0 | llama.cpp (Vulkan) | 190.78 | 7022.52 | – | – | tg128 | 2026-01-24 | community | – | source |
| RTX 4090 | llama 7B | Q4_0 | llama.cpp (Vulkan) | 190.1 | 10830.41 | – | – | tg128 | 2025-09-24 | community | – | source |
| GeForce RTX 4090 | llama 7B | Q4_0 | llama.cpp (CUDA) | 189.34 | 12415.31 | – | – | tg128 | 2025-08-24 | community | – | source |
| GeForce RTX 4090 | llama 7B | Q4_0 | llama.cpp (CUDA) | 189.33 | 10293.86 | – | – | tg128 | 2025-08-24 | community | – | source |
| RTX 4090 | llama 7B | Q4_0 | llama.cpp (CUDA) | 189.07 | 14753.2 | – | – | tg128 | 2025-08-03 | community | – | source |
| GeForce RTX 4090 | llama 7B | Q4_0 | llama.cpp (CUDA) | 188.96 | 14770.63 | – | – | tg128 | 2025-08-06 | community | – | source |
| RTX 4090 | llama 7B | Q4_0 | llama.cpp (Vulkan) | 187.97 | 9452.03 | – | – | tg128 | 2025-09-24 | community | – | source |
| GeForce RTX 4090 | llama 7B | Q4_0 | llama.cpp (CUDA) | 187.83 | 10259.68 | – | – | tg128 | 2025-08-24 | community | – | source |
| GeForce RTX 4090 | llama 7B | Q4_0 | llama.cpp (CUDA) | 186.21 | 11992.7 | – | – | tg128 | 2025-08-06 | community | – | source |
| A100 | llama 7B | Q4_0 | llama.cpp (CUDA) | 185.91 | 4883.69 | – | – | tg128 | 2025-09-05 | community | – | source |
| RTX 5080 | llama 7B | Q4_0 | llama.cpp (Vulkan) | 185.1 | 7444.99 | – | – | tg128 | 2026-01-31 | community | – | source |
| RTX 5080 | llama 7B | Q4_0 | llama.cpp (Vulkan) | 185.1 | 7444.99 | – | – | tg128 | 2026-01-31 | community | – | source |
| RTX 4090 | llama 7B | Q4_0 | llama.cpp (CUDA) | 185.04 | 14472.84 | – | – | tg128 | 2025-07-08 | community | – | source |
| GeForce RTX 5080 | llama 7B | Q4_0 | llama.cpp (CUDA) | 184.68 | 9487.7 | – | – | tg128 | 2025-08-28 | community | – | source |
| RTX 5080 | llama 7B | Q4_0 | llama.cpp (Vulkan) | 183.1 | 6840.43 | – | – | tg128 | 2026-01-24 | community | – | source |
| RTX 5080 | llama 7B | Q4_0 | llama.cpp (Vulkan) | 183.1 | 6840.43 | – | – | tg128 | 2026-01-24 | community | – | source |
| GeForce RTX 5080 | llama 7B | Q4_0 | llama.cpp (CUDA) | 183.04 | 9644.43 | – | – | tg128 | 2025-12-16 | community | – | source |
| Radeon RX 7900 XTX | llama 7B | Q4_0 | llama.cpp (Vulkan) | 182.63 | 3726.99 | – | – | tg128 | 2026-09-10 | community | – | source |
| GeForce RTX 5070 Ti | llama 7B | Q4_0 | llama.cpp (CUDA) | 182.43 | 8419.56 | – | – | tg128 | 2025-12-07 | community | – | source |
| GeForce RTX 5080 | llama 7B | Q4_0 | llama.cpp (CUDA) | 181.99 | 8297.36 | – | – | tg128 | 2025-08-28 | community | – | source |
| Radeon RX 7900 XTX | llama 7B | Q4_0 | llama.cpp (Vulkan) | 179.66 | 3932.57 | – | – | tg128 | 2026-08-23 | community | contradiction | source |
| RTX 4090 | llama 7B | Q4_0 | llama.cpp (Vulkan) | 179.62 | 9481.99 | – | – | tg128 | 2025-08-03 | community | – | source |
| NVIDIA RTX 6000 | llama 7B | Q4_0 | llama.cpp (CUDA) | 179.47 | 10576.85 | – | – | tg128 | 2025-09-22 | community | – | source |
| M5 Ultra | llama 7B 6.74B | Q4_0 | llama.cpp (Metal) | 179.1 | 4944.66 | – | – | tg128 | 2026-09-30 | community | – | source |
| GeForce RTX 5080 | llama 7B | Q4_0 | llama.cpp (CUDA) | 177.99 | 7405.11 | – | – | tg128 | 2025-12-16 | community | – | source |
| GeForce RTX 3090 | llama 7B | Q4_0 | llama.cpp (CUDA) | 177.79 | 5992.1 | – | – | tg128 | 2026-05-03 | community | – | source |
| GeForce RTX 3090 | llama 7B | Q4_0 | llama.cpp (CUDA) | 177.75 | 5973.72 | – | – | tg128 | 2026-09-07 | community | – | source |
| A100 | llama 7B | Q4_0 | llama.cpp (CUDA) | 176.88 | 5141.45 | – | – | tg128 | 2025-05-22 | community | – | source |
| GeForce RTX 5070 Ti | llama 7B | Q4_0 | llama.cpp (CUDA) | 176.85 | 6952.38 | – | – | tg128 | 2025-12-07 | community | – | source |
| Radeon 8060S | llama 7B | Q4_0 | llama.cpp (Vulkan) | 176.18 | 7567.94 | – | – | tg128 | 2026-04-24 | community | – | source |
| NVIDIA RTX 6000 | llama 7B | Q4_0 | llama.cpp (CUDA) | 176.07 | 9229.23 | – | – | tg128 | 2025-09-22 | community | – | source |
| RTX PRO 4500 Blackwell | llama 7B | Q4_0 | llama.cpp (CUDA) | 175.57 | 8149.86 | – | – | tg128 | 2025-11-14 | community | – | source |
| GeForce RTX 3090 | llama 7B | Q4_0 | llama.cpp (CUDA) | 175.48 | 5874.09 | – | – | tg128 | 2025-12-11 | community | – | source |
| Radeon RX 7900 XTX | llama 7B | Q4_0 | llama.cpp (Vulkan) | 173.81 | 3153.9 | – | – | tg128 | 2025-12-11 | community | – | source |
| Radeon RX 7900 XTX | llama 7B | Q4_0 | llama.cpp (Vulkan) | 173.73 | 3259.34 | – | – | tg128 | 2025-12-11 | community | – | source |
| Radeon RX 7900 XTX | llama 7B | Q4_0 | llama.cpp (Vulkan) | 173.36 | 3253.68 | – | – | tg128 | 2025-12-11 | community | – | source |
| Radeon RX 7900 XTX | llama 7B | Q4_0 | llama.cpp (Vulkan) | 173.13 | 3250.86 | – | – | tg128 | 2025-12-11 | community | – | source |
| Radeon RX 7900 XTX | llama 7B | Q4_0 | llama.cpp (Vulkan) | 172.86 | 3290.92 | – | – | tg128 | 2026-03-13 | community | – | source |
| Radeon RX 7900 XTX | llama 7B | Q4_0 | llama.cpp (Vulkan) | 172.84 | 2706.82 | – | – | tg128 | 2025-12-11 | community | – | source |
| Radeon RX 7900 XTX | llama 7B | Q4_0 | llama.cpp (Vulkan) | 172.79 | 2806.12 | – | – | tg128 | 2025-12-11 | community | – | source |
| Radeon RX 7900 XTX | llama 7B | Q4_0 | llama.cpp (Vulkan) | 172.76 | 2801.46 | – | – | tg128 | 2025-12-11 | community | – | source |
| Radeon RX 7900 XTX | llama 7B | Q4_0 | llama.cpp (Vulkan) | 172.71 | 2805.99 | – | – | tg128 | 2025-12-11 | community | – | source |
| GeForce RTX 3090 Ti | llama 7B | Q4_0 | llama.cpp (CUDA) | 172.26 | 6924.01 | – | – | tg128 | 2025-08-01 | community | – | source |
| Radeon 8060S | llama 7B | Q4_0 | llama.cpp (Vulkan) | 171.61 | 5200.68 | – | – | tg128 | 2026-05-03 | community | – | source |
| GeForce RTX 3090 Ti | llama 7B | Q4_0 | llama.cpp (CUDA) | 171.19 | 6567.49 | – | – | tg128 | 2025-08-01 | community | – | source |
| RTX 4090 | llama 7B | Q4_0 | llama.cpp (Vulkan) | 170.95 | 8534.56 | – | – | tg128 | 2025-07-08 | community | – | source |
| Radeon RX 7900 XTX | llama 7B | Q4_0 | llama.cpp (Vulkan) | 170.69 | 3692.87 | – | – | tg128 | 2026-08-23 | community | – | source |
| M4 | llama 7B | Q4_0 | llama.cpp (Vulkan) | 170.56 | 7064.4 | – | – | tg128 | 2025-12-17 | community | – | source |
| Radeon RX 7900 XTX | llama 7B | Q4_0 | llama.cpp (Vulkan) | 170.44 | 3374.48 | – | – | tg128 | 2026-03-18 | community | – | source |
| Radeon RX 7900 XTX | llama 7B | Q4_0 | llama.cpp (Vulkan) | 170.13 | 3178.79 | – | – | tg128 | 2025-11-20 | community | – | source |
| GeForce RTX 4090 | llama 7B | Q4_0 | llama.cpp (CUDA) | 170.12 | 14017.58 | – | – | tg128 | 2025-08-27 | community | – | source |
| Radeon RX 7900 XTX | llama 7B | Q4_0 | llama.cpp (ROCm) | 170.12 | 3874.25 | – | – | tg128 | 2025-11-05 | community | – | source |
| GeForce RTX 4090 | llama 7B | Q4_0 | llama.cpp (CUDA) | 169.74 | 12416.64 | – | – | tg128 | 2025-08-27 | community | – | source |
| GeForce RTX 4090 | llama 7B | Q4_0 | llama.cpp (CUDA) | 169.74 | 12416.64 | – | – | tg128 | 2025-08-27 | community | – | source |
| Radeon RX 7900 XTX | llama 7B | Q4_0 | llama.cpp (Vulkan) | 169.38 | 2994.51 | – | – | tg128 | 2025-12-11 | community | – | source |
| Radeon RX 7900 XTX | llama 7B | Q4_0 | llama.cpp (Vulkan) | 169.06 | 3138.42 | – | – | tg128 | 2025-12-11 | community | – | source |
| GeForce RTX 3090 | llama 7B | Q4_0 | llama.cpp (CUDA) | 168.95 | 5153.91 | – | – | tg128 | 2026-09-07 | community | – | source |
| RTX PRO 4500 Blackwell | llama 7B | Q4_0 | llama.cpp (CUDA) | 168.9 | 7251.66 | – | – | tg128 | 2025-11-14 | community | – | source |
| GeForce RTX 3090 | llama 7B | Q4_0 | llama.cpp (CUDA) | 168.83 | 5349.43 | – | – | tg128 | 2026-05-03 | community | – | source |
| Radeon RX 7900 XTX | llama 7B | Q4_0 | llama.cpp (Vulkan) | 168.77 | 3121.91 | – | – | tg128 | 2025-12-11 | community | – | source |
| Radeon RX 7900 XTX | llama 7B | Q4_0 | llama.cpp (Vulkan) | 168.74 | 2622.48 | – | – | tg128 | 2025-12-11 | community | – | source |
| Radeon RX 7900 XTX | llama 7B | Q4_0 | llama.cpp (Vulkan) | 168.63 | 3128.43 | – | – | tg128 | 2025-12-11 | community | – | source |
| Radeon RX 7900 XTX | llama 7B | Q4_0 | llama.cpp (Vulkan) | 168.62 | 3479.41 | – | – | tg128 | 2026-03-01 | community | – | source |
| Radeon RX 7900 XTX | llama 7B | Q4_0 | llama.cpp (Vulkan) | 168.53 | 2333.75 | – | – | tg128 | 2025-11-20 | community | – | source |
| Radeon RX 7900 XTX | llama 7B | Q4_0 | llama.cpp (Vulkan) | 168.53 | 2333.75 | – | – | tg128 | 2025-11-20 | community | – | source |
| A100 | llama 7B | Q4_0 | llama.cpp (CUDA) | 168.23 | 4865.98 | – | – | tg128 | 2025-05-22 | community | – | source |
| RTX 5070 Ti | llama 7B | Q4_0 | llama.cpp (CUDA) | 168.2 | 7802.37 | – | – | tg128 | 2025-05-10 | community | – | source |
| Radeon RX 7900 XTX | llama 7B | Q4_0 | llama.cpp (Vulkan) | 168.04 | 2714.98 | – | – | tg128 | 2025-12-11 | community | – | source |
| Radeon RX 7900 XTX | llama 7B | Q4_0 | llama.cpp (Vulkan) | 167.79 | 2716.65 | – | – | tg128 | 2025-12-11 | community | – | source |
| Radeon RX 7900 XTX | llama 7B | Q4_0 | llama.cpp (Vulkan) | 167.78 | 2727.7 | – | – | tg128 | 2025-12-11 | community | – | source |
| GeForce RTX 4090 | llama 7B | Q4_0 | llama.cpp (CUDA) | 167.53 | 10712.64 | – | – | tg128 | 2025-08-27 | community | – | source |
| GeForce RTX 4090 | llama 7B | Q4_0 | llama.cpp (CUDA) | 167.53 | 10712.64 | – | – | tg128 | 2025-08-27 | community | – | source |
| GeForce RTX 4090 | llama 7B | Q4_0 | llama.cpp (CUDA) | 167.52 | 11773.24 | – | – | tg128 | 2025-08-27 | community | – | source |
| Radeon RX 7900 XTX | llama 7B | Q4_0 | llama.cpp (Vulkan) | 167.42 | 3042.23 | – | – | tg128 | 2025-11-20 | community | – | source |
| GeForce RTX 3090 | llama 7B | Q4_0 | llama.cpp (CUDA) | 167.39 | 5134.61 | – | – | tg128 | 2025-12-11 | community | – | source |
| Radeon RX 7900 XTX | llama 7B | Q4_0 | llama.cpp (Vulkan) | 167.31 | 2175.84 | – | – | tg128 | 2025-11-20 | community | – | source |
| Radeon RX 7900 XTX | llama 7B | Q4_0 | llama.cpp (Vulkan) | 167.31 | 2175.84 | – | – | tg128 | 2025-11-20 | community | – | source |
| Radeon RX 7900 XTX | llama 7B | Q4_0 | llama.cpp (ROCm) | 167.11 | 3552.27 | – | – | tg128 | 2025-11-05 | community | – | source |
| RTX 4090 | llama 7B | Q4_0 | llama.cpp (Vulkan) | 166.45 | 9032.44 | – | – | tg128 | 2025-07-08 | community | – | source |
| Radeon RX 7900 XTX | llama 7B | Q4_0 | llama.cpp (Vulkan) | 166.25 | 2044.99 | – | – | tg128 | 2025-10-16 | community | – | source |
| Radeon RX 7900 XTX | llama 7B | Q4_0 | llama.cpp (Vulkan) | 165.94 | 3143.31 | – | – | tg128 | 2025-12-11 | community | – | source |
| Radeon RX 7900 XTX | llama 7B | Q4_0 | llama.cpp (Vulkan) | 165.57 | 3170.06 | – | – | tg128 | 2025-12-11 | community | – | source |
| Radeon RX 7900 XTX | llama 7B | Q4_0 | llama.cpp (Vulkan) | 165.38 | 3170.2 | – | – | tg128 | 2025-12-11 | community | – | source |
| Radeon RX 7900 XTX | llama 7B | Q4_0 | llama.cpp (Vulkan) | 165.38 | 2148.72 | – | – | tg128 | 2025-11-15 | community | – | source |
| Radeon RX 7900 XTX | llama 7B | Q4_0 | llama.cpp (Vulkan) | 165.23 | 3165.77 | – | – | tg128 | 2025-12-11 | community | – | source |
| Radeon RX 7900 XTX | llama 7B | Q4_0 | llama.cpp (Vulkan) | 165.08 | 2281.62 | – | – | tg128 | 2025-10-16 | community | – | source |
| Radeon RX 7900 XTX | llama 7B | Q4_0 | llama.cpp (Vulkan) | 164.84 | 2750.74 | – | – | tg128 | 2025-12-11 | community | – | source |
| Radeon RX 7900 XTX | llama 7B | Q4_0 | llama.cpp (Vulkan) | 164.32 | 2805.52 | – | – | tg128 | 2025-12-11 | community | – | source |
| GeForce RTX 5090 | llama 7B | Q4_0 | llama.cpp (CUDA) | 164.31 | 7603.2 | – | – | tg128 | 2026-09-16 | community | – | source |
| Radeon RX 7900 XTX | llama 7B | Q4_0 | llama.cpp (Vulkan) | 164.28 | 3181.81 | – | – | tg128 | 2025-11-20 | community | – | source |
| Radeon RX 7900 XTX | llama 7B | Q4_0 | llama.cpp (Vulkan) | 164.22 | 2814.04 | – | – | tg128 | 2025-12-11 | community | – | source |
| Radeon RX 7900 XTX | llama 7B | Q4_0 | llama.cpp (Vulkan) | 164.21 | 2807.29 | – | – | tg128 | 2025-12-11 | community | – | source |
| Radeon 8060S | llama 7B | Q4_0 | llama.cpp (Vulkan) | 164.05 | 4666.15 | – | – | tg128 | 2026-05-03 | community | – | source |
| Radeon RX 7900 XTX | llama 7B | Q4_0 | llama.cpp (ROCm) | 163.77 | 4403.4 | – | – | tg128 | 2026-09-10 | community | – | source |
| RTX 3090 | llama 7B | Q4_0 | llama.cpp (CUDA) | 163.69 | 5781.6 | – | – | tg128 | 2025-09-01 | community | – | source |
| Radeon RX 7900 XTX | llama 7B | Q4_0 | llama.cpp (Vulkan) | 163.0 | 1997.19 | – | – | tg128 | 2025-10-16 | community | – | source |
| Radeon RX 7900 XTX | llama 7B | Q4_0 | llama.cpp (Vulkan) | 162.78 | 2284.71 | – | – | tg128 | 2025-11-20 | community | – | source |
| Radeon RX 7900 XTX | llama 7B | Q4_0 | llama.cpp (Vulkan) | 162.78 | 2284.71 | – | – | tg128 | 2025-11-20 | community | – | source |
| RTX 3090 | llama 7B | Q4_0 | llama.cpp (Vulkan) | 162.28 | 4732.33 | – | – | tg128 | 2025-09-24 | community | – | source |
| Radeon RX 7900 XTX | llama 7B | Q4_0 | llama.cpp (Vulkan) | 162.16 | 2254.0 | – | – | tg128 | 2025-10-16 | community | – | source |
| Radeon RX 7900 XTX | llama 7B | Q4_0 | llama.cpp (Vulkan) | 162.02 | 2170.71 | – | – | tg128 | 2025-11-20 | community | – | source |
| Radeon RX 7900 XTX | llama 7B | Q4_0 | llama.cpp (Vulkan) | 162.02 | 2170.71 | – | – | tg128 | 2025-11-20 | community | – | source |
| Radeon RX 7900 XTX | llama 7B | Q4_0 | llama.cpp (Vulkan) | 161.63 | 3044.63 | – | – | tg128 | 2025-11-20 | community | – | source |
| Radeon RX 7900 XTX | llama 7B | Q4_0 | llama.cpp (Vulkan) | 161.53 | 3164.42 | – | – | tg128 | 2026-03-18 | community | – | source |
| Radeon RX 7900 XTX | llama 7B | Q4_0 | llama.cpp (Vulkan) | 161.48 | 2973.55 | – | – | tg128 | 2025-12-11 | community | – | source |
| Radeon RX 7900 XTX | llama 7B | Q4_0 | llama.cpp (Vulkan) | 161.46 | 3453.79 | – | – | tg128 | 2026-02-16 | community | contradiction | source |
| Radeon RX 7900 XTX | llama 7B | Q4_0 | llama.cpp (Vulkan) | 161.39 | 3069.84 | – | – | tg128 | 2025-12-11 | community | – | source |
| Radeon RX 7900 XTX | llama 7B | Q4_0 | llama.cpp (Vulkan) | 161.07 | 3338.82 | – | – | tg128 | 2026-03-01 | community | – | source |
| Radeon RX 7900 XTX | llama 7B | Q4_0 | llama.cpp (Vulkan) | 160.78 | 3051.47 | – | – | tg128 | 2025-12-11 | community | – | source |
| M4 | llama 7B | Q4_0 | llama.cpp (Vulkan) | 160.78 | 6389.86 | – | – | tg128 | 2025-12-17 | community | – | source |
| Radeon RX 7900 XTX | llama 7B | Q4_0 | llama.cpp (Vulkan) | 160.77 | 3056.15 | – | – | tg128 | 2025-12-11 | community | – | source |
| Radeon RX 7900 XTX | llama 7B | Q4_0 | llama.cpp (Vulkan) | 160.77 | 2653.13 | – | – | tg128 | 2025-12-11 | community | – | source |
| GeForce RTX 3090 | llama 7B | Q4_0 | llama.cpp (CUDA) | 160.67 | 5051.48 | – | – | tg128 | 2025-12-11 | community | – | source |
| Radeon RX 7900 XTX | llama 7B | Q4_0 | llama.cpp (Vulkan) | 160.2 | 2739.28 | – | – | tg128 | 2025-12-11 | community | – | source |
| RTX 3090 | llama 7B | Q4_0 | llama.cpp (Vulkan) | 160.13 | 4298.97 | – | – | tg128 | 2025-09-24 | community | – | source |
| Radeon RX 7900 XTX | llama 7B | Q4_0 | llama.cpp (Vulkan) | 159.86 | 2727.9 | – | – | tg128 | 2025-12-11 | community | – | source |
| Radeon RX 7900 XTX | llama 7B | Q4_0 | llama.cpp (Vulkan) | 159.78 | 2726.36 | – | – | tg128 | 2025-12-11 | community | – | source |
| Radeon RX 7900 XTX | llama 7B | Q4_0 | llama.cpp (Vulkan) | 159.46 | 2162.71 | – | – | tg128 | 2025-11-15 | community | – | source |
| RTX 3090 | llama 7B | Q4_0 | llama.cpp (CUDA) | 159.36 | 5100.68 | – | – | tg128 | 2025-09-01 | community | – | source |
| GeForce RTX 5090 | llama 7B | Q4_0 | llama.cpp (CUDA) | 158.14 | 7641.89 | – | – | tg128 | 2026-05-26 | community | – | source |
| Radeon RX 7900 XTX | llama 7B | Q4_0 | llama.cpp (ROCm) | 157.91 | 3266.07 | – | – | tg128 | 2025-02-06 | community | – | source |
| GeForce RTX 5090 | llama 7B | Q4_0 | llama.cpp (CUDA) | 157.84 | 6663.79 | 150.0 | – | tg128 | 2026-09-16 | community | – | source |
| RTX 3090 | llama 7B | Q4_0 | llama.cpp (Vulkan) | 157.82 | 4409.02 | – | – | tg128 | 2025-09-01 | community | – | source |
| Radeon RX 7900 XTX | llama 7B | Q4_0 | llama.cpp (Vulkan) | 157.41 | 3422.83 | – | – | tg128 | 2026-08-23 | community | – | source |
| RTX 3090 | llama 7B | Q4_0 | llama.cpp (Vulkan) | 156.57 | 4892.33 | – | – | tg128 | 2025-09-01 | community | – | source |
| GeForce RTX 5090 | llama 7B | Q4_0 | llama.cpp (CUDA) | 156.49 | 6667.06 | – | – | tg128 | 2026-05-26 | community | – | source |
| A100 | llama 7B | Q4_0 | llama.cpp (CUDA) | 155.72 | 3016.29 | – | – | tg128 | 2026-07-30 | community | – | source |
| Radeon RX 7900 XTX | llama 7B | Q4_0 | llama.cpp (ROCm) | 155.69 | 4145.55 | – | – | tg128 | 2026-09-10 | community | – | source |
| GeForce RTX 3090 | llama 7B | Q4_0 | llama.cpp (CUDA) | 155.36 | 5932.44 | – | – | tg128 | 2025-08-02 | community | – | source |
| Radeon RX 7900 XTX | llama 7B | Q4_0 | llama.cpp (Vulkan) | 154.32 | 3292.37 | – | – | tg128 | 2026-02-16 | community | – | source |
| RTX 5090 | llama 7B | Q4_0 | llama.cpp (Vulkan) | 154.31 | 6011.18 | 160.0 | – | tg128 | 2026-09-17 | community | – | source |
| Radeon RX 7900 XTX | llama 7B | Q4_0 | llama.cpp (ROCm) | 153.91 | 3434.01 | – | – | tg128 | 2025-08-06 | community | – | source |
| Radeon RX 7900 XTX | llama 7B | Q4_0 | llama.cpp (ROCm) | 153.88 | 4190.07 | – | – | tg128 | 2026-08-23 | community | – | source |
| Radeon RX 7900 XTX | llama 7B | Q4_0 | llama.cpp (ROCm) | 153.72 | 3599.57 | – | – | tg128 | 2025-05-26 | community | – | source |
| RTX 5070 Ti | llama 7B | Q4_0 | llama.cpp (CUDA) | 153.55 | 7623.55 | – | – | tg128 | 2025-05-23 | community | – | source |
| GeForce RTX 3090 | llama 7B | Q4_0 | llama.cpp (CUDA) | 153.01 | 4425.82 | – | – | tg128 | 2025-12-11 | community | – | source |
| V100 | llama 7B | Q4_0 | llama.cpp (CUDA) | 152.36 | 3134.81 | – | – | tg128 | 2025-11-19 | community | – | source |
| V100 | llama 7B | Q4_0 | llama.cpp (CUDA) | 152.33 | 3537.71 | – | – | tg128 | 2025-12-17 | community | – | source |
| V100 | llama 7B | Q4_0 | llama.cpp (CUDA) | 151.88 | 3133.71 | – | – | tg128 | 2025-11-19 | community | – | source |
| RTX 3090 | llama 7B | Q4_0 | llama.cpp (Vulkan) | 151.65 | 4356.94 | – | – | tg128 | 2025-12-11 | community | – | source |
| RTX 3090 | llama 7B | Q4_0 | llama.cpp (Vulkan) | 151.65 | 4435.33 | – | – | tg128 | 2025-12-11 | community | – | source |
| GeForce RTX 3090 | llama 7B | Q4_0 | llama.cpp (CUDA) | 151.04 | 5405.83 | – | – | tg128 | 2025-08-02 | community | – | source |
| Radeon RX 7900 XTX | llama 7B | Q4_0 | llama.cpp (Vulkan) | 150.97 | 3229.07 | – | – | tg128 | 2026-03-13 | community | contradiction | source |
| Radeon RX 7900 XTX | llama 7B | Q4_0 | llama.cpp (ROCm) | 150.86 | 3529.1 | – | – | tg128 | 2025-07-22 | community | – | source |
| V100 | llama 7B | Q4_0 | llama.cpp (CUDA) | 150.68 | 3075.88 | – | – | tg128 | 2025-11-19 | community | – | source |
| Radeon RX 7900 XTX | llama 7B | Q4_0 | llama.cpp (ROCm) | 150.4 | 3458.31 | – | – | tg128 | 2026-03-18 | community | – | source |
| Radeon RX 7900 XTX | llama 7B | Q4_0 | llama.cpp (Vulkan) | 150.34 | 2369.54 | – | – | tg128 | 2025-11-15 | community | contradiction | source |
| RTX 4080 SUPER | llama 7B | Q4_0 | llama.cpp (Vulkan) | 150.2 | 8007.37 | – | – | tg128 | 2025-10-18 | community | – | source |
| RTX 5090 | llama 7B | Q4_0 | llama.cpp (Vulkan) | 149.41 | 5546.39 | – | – | tg128 | 2026-09-17 | community | – | source |
| RTX 5070 Ti | llama 7B | Q4_0 | llama.cpp (CUDA) | 149.28 | 6870.08 | – | – | tg128 | 2025-05-23 | community | – | source |
| Radeon RX 7900 XTX | llama 7B | Q4_0 | llama.cpp (Vulkan) | 149.04 | 3252.09 | – | – | tg128 | 2026-08-23 | community | – | source |
| Radeon RX 7900 XTX | llama 7B | Q4_0 | llama.cpp (Vulkan) | 148.34 | 3134.2 | – | – | tg128 | 2025-08-06 | community | – | source |
| GeForce RTX 4080 Super | llama 7B | Q4_0 | llama.cpp (CUDA) | 148.33 | 8125.15 | – | – | tg128 | 2025-10-18 | community | – | source |
| Radeon RX 7900 XTX | llama 7B | Q4_0 | llama.cpp (Vulkan) | 148.21 | 3236.63 | – | – | tg128 | 2025-02-06 | community | – | source |
| Radeon RX 7900 XTX | llama 7B | Q4_0 | llama.cpp (Vulkan) | 148.01 | 3288.72 | – | – | tg128 | 2025-08-06 | community | – | source |
| GeForce RTX 4080 Super | llama 7B | Q4_0 | llama.cpp (CUDA) | 147.48 | 9439.01 | – | – | tg128 | 2025-10-18 | community | – | source |
| Radeon RX 7900 XTX | llama 7B | Q4_0 | llama.cpp (Vulkan) | 147.25 | 1943.62 | – | – | tg128 | 2025-06-23 | community | – | source |
| RTX 4080 SUPER | llama 7B | Q4_0 | llama.cpp (Vulkan) | 147.13 | 7101.18 | – | – | tg128 | 2025-10-18 | community | – | source |
| Radeon RX 7900 XTX | llama 7B | Q4_0 | llama.cpp (ROCm) | 147.06 | 3875.83 | – | – | tg128 | 2026-08-23 | community | – | source |
| Radeon RX 7900 XTX | llama 7B | Q4_0 | llama.cpp (Vulkan) | 146.66 | 1986.43 | – | – | tg128 | 2025-05-20 | community | – | source |
| Radeon RX 7900 XTX | llama 7B | Q4_0 | llama.cpp (Vulkan) | 145.78 | 1952.7 | – | – | tg128 | 2025-06-23 | community | – | source |
| Radeon RX 7900 XTX | llama 7B | Q4_0 | llama.cpp (Vulkan) | 145.7 | 3225.19 | – | – | tg128 | 2025-07-22 | community | – | source |
| V100 | llama 7B | Q4_0 | llama.cpp (CUDA) | 145.45 | 3434.83 | – | – | tg128 | 2025-12-17 | community | – | source |
| Radeon RX 7900 XTX | llama 7B | Q4_0 | llama.cpp (Vulkan) | 145.35 | 1937.34 | – | – | tg128 | 2025-07-17 | community | – | source |
| Radeon RX 7900 XTX | llama 7B | Q4_0 | llama.cpp (ROCm) | 145.32 | 3966.95 | – | – | tg128 | 2026-03-18 | community | – | source |
| Radeon RX 7900 XTX | llama 7B | Q4_0 | llama.cpp (ROCm) | 145.23 | 3633.86 | – | – | tg128 | 2025-08-06 | community | – | source |
| Radeon RX 7900 XTX | llama 7B | Q4_0 | llama.cpp (Vulkan) | 145.0 | 3489.67 | – | – | tg128 | 2025-07-22 | community | – | source |
| Radeon RX 7900 XTX | llama 7B | Q4_0 | llama.cpp (Vulkan) | 144.9 | 3419.13 | – | – | tg128 | 2025-05-26 | community | – | source |
| A6000 | llama 7B | Q4_0 | llama.cpp (CUDA) | 144.87 | 5662.39 | – | – | tg128 | 2025-09-01 | community | – | source |
| Radeon RX 7900 XTX | llama 7B | Q4_0 | llama.cpp (ROCm) | 144.84 | 3552.47 | – | – | tg128 | 2026-03-18 | community | – | source |
| Radeon RX 7900 XTX | llama 7B | Q4_0 | llama.cpp (Vulkan) | 143.99 | 2062.17 | – | – | tg128 | 2025-01-08 | community | – | source |
| Radeon RX 7900 XTX | llama 7B | Q4_0 | llama.cpp (ROCm) | 143.98 | 3743.15 | – | – | tg128 | 2025-05-26 | community | – | source |
| GeForce RTX 3080 | llama 7B | Q4_0 | llama.cpp (CUDA) | 143.88 | 5438.27 | – | – | tg128 | 2026-04-03 | community | – | source |
| Radeon RX 7900 XTX | llama 7B | Q4_0 | llama.cpp (Vulkan) | 143.75 | 2222.31 | – | – | tg128 | 2025-11-15 | community | contradiction | source |
| GeForce RTX 4080 | llama 7B | Q4_0 | llama.cpp (CUDA) | 143.47 | 9205.93 | – | – | tg128 | 2025-08-07 | community | – | source |
| RTX 3090 | llama 7B | Q4_0 | llama.cpp (Vulkan) | 143.46 | 3877.12 | – | – | tg128 | 2025-12-11 | community | – | source |
| RTX 3080 | llama 7B | Q4_0 | llama.cpp (Vulkan) | 143.44 | 4889.13 | – | – | tg128 | 2026-04-03 | community | – | source |
| Radeon RX 9070 XT | llama 7B | Q4_0 | llama.cpp (Vulkan) | 142.88 | 6004.06 | – | – | tg128 | 2026-06-05 | community | – | source |
| RTX 3090 | llama 7B | Q4_0 | llama.cpp (Vulkan) | 142.76 | 4033.29 | – | – | tg128 | 2025-12-11 | community | – | source |
| GeForce RTX 4080 | llama 7B | Q4_0 | llama.cpp (CUDA) | 142.49 | 8031.64 | – | – | tg128 | 2025-08-07 | community | – | source |
| Radeon RX 7900 XTX | llama 7B | Q4_0 | llama.cpp (ROCm) | 142.17 | 3764.9 | – | – | tg128 | 2026-08-23 | community | – | source |
| M2 | llama 7B | Q4_0 | llama.cpp (Vulkan) | 142.13 | 1411.25 | – | – | tg128 | 2025-11-19 | community | – | source |
| Radeon RX 7900 XTX | llama 7B | Q4_0 | llama.cpp (Vulkan) | 141.52 | 3187.61 | – | – | tg128 | 2025-06-13 | community | – | source |
| A100 | llama 7B | Q4_0 | llama.cpp (CUDA) | 141.32 | 2698.95 | – | – | tg128 | 2026-07-30 | community | – | source |
| Radeon RX 7900 XTX | llama 7B | Q4_0 | llama.cpp (ROCm) | 140.96 | 3598.42 | – | – | tg128 | 2025-07-22 | community | – | source |
| Radeon RX 7900 XTX | llama 7B | Q4_0 | llama.cpp (ROCm) | 140.94 | 3516.41 | – | – | tg128 | 2025-11-20 | community | – | source |
| Radeon RX 7900 XTX | llama 7B | Q4_0 | llama.cpp (ROCm) | 140.94 | 3516.41 | – | – | tg128 | 2025-11-20 | community | – | source |
| Radeon RX 7900 XTX | llama 7B | Q4_0 | llama.cpp (ROCm) | 140.83 | 3577.58 | – | – | tg128 | 2025-11-20 | community | – | source |
| Radeon RX 7900 XTX | llama 7B | Q4_0 | llama.cpp (Vulkan) | 140.76 | 2001.05 | – | – | tg128 | 2025-05-20 | community | – | source |
| Radeon RX 7900 XTX | llama 7B | Q4_0 | llama.cpp (Vulkan) | 140.68 | 2039.24 | – | – | tg128 | 2025-01-08 | community | – | source |
| A5000 | llama 7B | Q4_0 | llama.cpp (Vulkan) | 140.43 | 4071.22 | – | – | tg128 | 2025-09-24 | community | – | source |
| Radeon RX 7900 XTX | llama 7B | Q4_0 | llama.cpp (ROCm) | 140.27 | 3588.15 | – | – | tg128 | 2025-11-20 | community | – | source |
| Radeon RX 7900 XTX | llama 7B | Q4_0 | llama.cpp (ROCm) | 140.27 | 3588.15 | – | – | tg128 | 2025-11-20 | community | – | source |
| V100 | llama 7B | Q4_0 | llama.cpp (CUDA) | 140.23 | 3261.46 | – | – | tg128 | 2025-12-17 | community | – | source |
| V100 | llama 7B | Q4_0 | llama.cpp (CUDA) | 140.0 | 3235.72 | – | – | tg128 | 2025-11-19 | community | – | source |
| GeForce RTX 3080 | llama 7B | Q4_0 | llama.cpp (CUDA) | 139.95 | 5569.56 | – | – | tg128 | 2025-08-01 | community | – | source |
| A5000 | llama 7B | Q4_0 | llama.cpp (Vulkan) | 139.89 | 3641.55 | – | – | tg128 | 2025-09-24 | community | – | source |
| V100 | llama 7B | Q4_0 | llama.cpp (CUDA) | 139.89 | 3235.01 | – | – | tg128 | 2025-11-19 | community | – | source |
| Radeon RX 7900 XTX | llama 7B | Q4_0 | llama.cpp (ROCm) | 139.89 | 3623.53 | – | – | tg128 | 2026-03-18 | community | – | source |
| Radeon RX 7900 XTX | llama 7B | Q4_0 | llama.cpp (Vulkan) | 139.81 | 3321.11 | – | – | tg128 | 2026-03-20 | community | contradiction | source |
| Radeon RX 7900 XTX | llama 7B | Q4_0 | llama.cpp (ROCm) | 139.68 | 3482.15 | – | – | tg128 | 2025-11-20 | community | – | source |
| GeForce RTX 3080 | llama 7B | Q4_0 | llama.cpp (CUDA) | 139.65 | 5013.86 | – | – | tg128 | 2025-08-01 | community | – | source |
| V100 | llama 7B | Q4_0 | llama.cpp (CUDA) | 139.11 | 3149.71 | – | – | tg128 | 2025-11-19 | community | – | source |
| Radeon RX 7900 XTX | llama 7B | Q4_0 | llama.cpp (ROCm) | 138.91 | 3782.88 | – | – | tg128 | 2025-12-10 | community | – | source |
| Radeon RX 9070 XT | llama 7B | Q4_0 | llama.cpp (Vulkan) | 138.84 | 3895.62 | – | – | tg128 | 2025-09-23 | community | – | source |
| A6000 | llama 7B | Q4_0 | llama.cpp (CUDA) | 138.73 | 4913.93 | – | – | tg128 | 2025-09-01 | community | – | source |
| Radeon RX 9070 XT | llama 7B | Q4_0 | llama.cpp (Vulkan) | 138.45 | 3525.21 | – | – | tg128 | 2025-09-23 | community | – | source |
| Radeon RX 7900 XT | llama 7B | Q4_0 | llama.cpp (Vulkan) | 137.78 | 1619.74 | – | – | tg128 | 2025-08-23 | community | – | source |
| Radeon RX 9070 XT | llama 7B | Q4_0 | llama.cpp (Vulkan) | 137.31 | 4262.85 | – | – | tg128 | 2025-11-29 | community | – | source |
| Radeon RX 9070 XT | llama 7B | Q4_0 | llama.cpp (Vulkan) | 137.11 | 5036.04 | – | – | tg128 | 2026-01-24 | community | – | source |
| Radeon RX 7900 XTX | llama 7B | Q4_0 | llama.cpp (Vulkan) | 136.98 | 1997.04 | – | – | tg128 | 2025-01-08 | community | – | source |
| RTX 3090 | llama 7B | Q4_0 | llama.cpp (Vulkan) | 136.81 | 4543.96 | – | – | tg128 | 2025-07-08 | community | – | source |
| Radeon RX 9070 XT | llama 7B | Q4_0 | llama.cpp (Vulkan) | 136.64 | 5673.65 | – | – | tg128 | 2026-06-05 | community | – | source |
| RTX PRO 4000 Blackwell | llama 7B | Q4_0 | llama.cpp (CUDA) | 136.38 | 5674.44 | – | – | tg128 | 2025-11-21 | community | – | source |
| Radeon RX 7900 XTX | llama 7B | Q4_0 | llama.cpp (Vulkan) | 136.24 | 2022.59 | – | – | tg128 | 2025-01-08 | community | – | source |
| Radeon RX 7900 XTX | llama 7B | Q4_0 | llama.cpp (ROCm) | 136.05 | 3537.94 | – | – | tg128 | 2026-08-23 | community | – | source |
| A5000 | llama 7B | Q4_0 | llama.cpp (CUDA) | 135.83 | 4552.15 | – | – | tg128 | 2025-08-19 | community | – | source |
| RTX 4070 Ti SUPER | llama 7B | Q4_0 | llama.cpp (Vulkan) | 135.81 | 6801.18 | – | – | tg128 | 2025-09-24 | community | – | source |
| Radeon RX 7900 XT | llama 7B | Q4_0 | llama.cpp (Vulkan) | 135.65 | 1619.49 | – | – | tg128 | 2025-08-23 | community | – | source |
| RTX 5070 Ti | llama 7B | Q4_0 | llama.cpp (Vulkan) | 135.63 | 6213.63 | – | – | tg128 | 2025-05-23 | community | – | source |
| GeForce RTX 3080 | llama 7B | Q4_0 | llama.cpp (CUDA) | 135.07 | 4901.77 | – | – | tg128 | 2025-08-23 | community | – | source |
| Radeon RX 7900 XTX | llama 7B | Q4_0 | llama.cpp (Vulkan) | 134.87 | 1151.72 | – | – | tg128 | 2026-03-01 | community | contradiction | source |
| V100 | llama 7B | Q4_0 | llama.cpp (CUDA) | 134.76 | 2973.78 | – | – | tg128 | 2025-09-22 | community | – | source |
| Radeon RX 7900 XTX | llama 7B | Q4_0 | llama.cpp (ROCm) | 134.46 | 1155.75 | – | – | tg128 | 2026-02-16 | community | – | source |
| Radeon RX 7900 XTX | llama 7B | Q4_0 | llama.cpp (Vulkan) | 134.46 | 3245.11 | – | – | tg128 | 2026-03-20 | community | contradiction | source |
| Radeon RX 7900 XTX | llama 7B | Q4_0 | llama.cpp (ROCm) | 134.07 | 3309.54 | – | – | tg128 | 2025-11-20 | community | – | source |
| Radeon RX 7900 XTX | llama 7B | Q4_0 | llama.cpp (ROCm) | 134.07 | 3309.54 | – | – | tg128 | 2025-11-20 | community | – | source |
| Radeon RX 7900 XTX | llama 7B | Q4_0 | llama.cpp (ROCm) | 134.01 | 3364.06 | – | – | tg128 | 2025-11-20 | community | – | source |
| RTX 5070 Ti | llama 7B | Q4_0 | llama.cpp (Vulkan) | 133.94 | 6614.86 | – | – | tg128 | 2025-05-23 | community | – | source |
| RTX 3090 | llama 7B | Q4_0 | llama.cpp (Vulkan) | 133.63 | 4946.47 | – | – | tg128 | 2025-07-08 | community | – | source |
| Radeon RX 7900 XTX | llama 7B | Q4_0 | llama.cpp (ROCm) | 133.59 | 3365.22 | – | – | tg128 | 2025-11-20 | community | – | source |
| Radeon RX 7900 XTX | llama 7B | Q4_0 | llama.cpp (ROCm) | 133.59 | 3365.22 | – | – | tg128 | 2025-11-20 | community | – | source |
| Radeon RX 7900 XTX | llama 7B | Q4_0 | llama.cpp (Vulkan) | 133.42 | 3110.71 | – | – | tg128 | 2025-06-23 | community | – | source |
| V100 | llama 7B | Q4_0 | llama.cpp (CUDA) | 133.06 | 3160.43 | – | – | tg128 | 2025-12-17 | community | – | source |
| Radeon RX 7900 XTX | llama 7B | Q4_0 | llama.cpp (Vulkan) | 133.02 | 1876.28 | – | – | tg128 | 2025-07-17 | community | – | source |
| GeForce RTX 4070 Ti Super | llama 7B | Q4_0 | llama.cpp (CUDA) | 132.85 | 7612.32 | – | – | tg128 | 2025-08-01 | community | – | source |
| Radeon RX 7900 XTX | llama 7B | Q4_0 | llama.cpp (ROCm) | 132.78 | 3278.06 | – | – | tg128 | 2025-11-20 | community | – | source |
| Radeon RX 7900 XTX | llama 7B | Q4_0 | llama.cpp (Vulkan) | 132.75 | 3085.75 | – | – | tg128 | 2025-06-28 | community | – | source |
| Radeon RX 7900 XTX | llama 7B | Q4_0 | llama.cpp (ROCm) | 132.75 | 3316.94 | – | – | tg128 | 2026-08-23 | community | – | source |
| Radeon 880M | llama 7B | Q4_0 | llama.cpp (ROCm) | 132.63 | 2727.83 | – | – | tg128 | 2025-10-28 | community | – | source |
| Radeon RX 9070 XT | llama 7B | Q4_0 | llama.cpp (Vulkan) | 132.27 | 3465.31 | – | – | tg128 | 2025-12-18 | community | – | source |
| GeForce RTX 4070 Ti Super | llama 7B | Q4_0 | llama.cpp (CUDA) | 132.26 | 6924.53 | – | – | tg128 | 2025-08-01 | community | – | source |
| GeForce RTX 3080 | llama 7B | Q4_0 | llama.cpp (CUDA) | 132.18 | 4293.66 | – | – | tg128 | 2025-08-23 | community | – | source |
| Radeon RX 7900 XTX | llama 7B | Q4_0 | llama.cpp (ROCm) | 132.17 | 3473.24 | – | – | tg128 | 2025-08-13 | community | – | source |
| Radeon 880M | llama 7B | Q4_0 | llama.cpp (ROCm) | 132.12 | 2632.74 | – | – | tg128 | 2025-10-28 | community | – | source |
| Radeon RX 9070 XT | llama 7B | Q4_0 | llama.cpp (Vulkan) | 131.99 | 4273.68 | – | – | tg128 | 2025-11-29 | community | – | source |
| Radeon RX 7900 XTX | llama 7B | Q4_0 | llama.cpp (Vulkan) | 131.95 | 3224.31 | – | – | tg128 | 2025-07-17 | community | – | source |
| Radeon RX 7900 XTX | llama 7B | Q4_0 | llama.cpp (ROCm) | 131.7 | 3541.67 | – | – | tg128 | 2025-12-10 | community | – | source |
| RTX PRO 4000 Blackwell | llama 7B | Q4_0 | llama.cpp (CUDA) | 131.66 | 4992.83 | – | – | tg128 | 2025-11-21 | community | – | source |
| Radeon RX 9070 XT | llama 7B | Q4_0 | llama.cpp (Vulkan) | 131.54 | 5048.07 | – | – | tg128 | 2026-01-24 | community | – | source |
| Radeon AI PRO R9700 | llama 7B | Q4_0 | llama.cpp (Vulkan) | 131.19 | 5411.28 | – | – | tg128 | 2026-09-12 | community | – | source |
| RTX 3080 | llama 7B | Q4_0 | llama.cpp (CUDA) | 131.01 | 4499.47 | – | – | tg128 | 2024-12-18 | community | – | source |
| Radeon RX 9070 XT | llama 7B | Q4_0 | llama.cpp (Vulkan) | 131.01 | 5505.31 | – | – | tg128 | 2026-03-27 | community | – | source |
| Radeon AI PRO R9700 | llama 7B | Q4_0 | llama.cpp (Vulkan) | 130.97 | 4235.87 | – | – | tg128 | 2026-09-12 | community | – | source |
| Radeon RX 7900 XTX | llama 7B | Q4_0 | llama.cpp (Vulkan) | 130.9 | 2925.56 | – | – | tg128 | 2025-06-23 | community | – | source |
| Radeon RX 9070 XT | llama 7B | Q4_0 | llama.cpp (Vulkan) | 130.57 | 3831.64 | – | – | tg128 | 2025-08-07 | community | – | source |
| Radeon RX 9070 XT | llama 7B | Q4_0 | llama.cpp (Vulkan) | 130.56 | 3461.64 | – | – | tg128 | 2025-08-07 | community | – | source |
| Radeon RX 9070 XT | llama 7B | Q4_0 | llama.cpp (Vulkan) | 130.12 | 1964.92 | – | – | tg128 | 2025-08-24 | community | – | source |
| A5000 | llama 7B | Q4_0 | llama.cpp (CUDA) | 130.07 | 4028.16 | – | – | tg128 | 2025-08-19 | community | – | source |
| Radeon RX 7900 XTX | llama 7B | Q4_0 | llama.cpp (Vulkan) | 129.8 | 2546.13 | – | – | tg128 | 2026-03-13 | community | contradiction | source |
| Radeon RX 7900 XTX | llama 7B | Q4_0 | llama.cpp (Vulkan) | 129.69 | 1313.1 | – | – | tg128 | 2026-03-01 | community | contradiction | source |
| M2 | llama 7B | Q4_0 | llama.cpp (Vulkan) | 129.58 | 1391.39 | – | – | tg128 | 2025-11-19 | community | – | source |
| Radeon RX 7900 XTX | llama 7B | Q4_0 | llama.cpp (Vulkan) | 129.48 | 2281.29 | – | – | tg128 | 2025-12-11 | community | – | source |
| RTX 4070 Ti SUPER | llama 7B | Q4_0 | llama.cpp (Vulkan) | 129.45 | 6099.18 | – | – | tg128 | 2025-09-24 | community | – | source |
| Radeon RX 7900 XTX | llama 7B | Q4_0 | llama.cpp (ROCm) | 129.33 | 1253.3 | – | – | tg128 | 2026-02-16 | community | – | source |
| Radeon RX 7900 XTX | llama 7B | Q4_0 | llama.cpp (Vulkan) | 129.3 | 2285.41 | – | – | tg128 | 2025-12-11 | community | – | source |
| Radeon RX 7900 XTX | llama 7B | Q4_0 | llama.cpp (Vulkan) | 129.22 | 2561.41 | – | – | tg128 | 2025-11-20 | community | contradiction | source |
| Radeon RX 7900 XTX | llama 7B | Q4_0 | llama.cpp (Vulkan) | 129.2 | 1878.23 | – | – | tg128 | 2025-12-11 | community | – | source |
| Radeon RX 7900 XTX | llama 7B | Q4_0 | llama.cpp (Vulkan) | 129.15 | 2278.5 | – | – | tg128 | 2025-12-11 | community | – | source |
| V100 | llama 7B | Q4_0 | llama.cpp (CUDA) | 129.08 | 3042.64 | – | – | tg128 | 2025-09-22 | community | – | source |
| Radeon RX 7900 XTX | llama 7B | Q4_0 | llama.cpp (Vulkan) | 129.07 | 2130.63 | – | – | tg128 | 2025-12-11 | community | – | source |
| Radeon RX 7900 XTX | llama 7B | Q4_0 | llama.cpp (Vulkan) | 129.07 | 2590.43 | – | – | tg128 | 2025-12-11 | community | – | source |
| Radeon RX 7900 XTX | llama 7B | Q4_0 | llama.cpp (Vulkan) | 128.95 | 2580.6 | – | – | tg128 | 2025-12-11 | community | – | source |
| Radeon RX 7900 XTX | llama 7B | Q4_0 | llama.cpp (Vulkan) | 128.94 | 2581.45 | – | – | tg128 | 2025-12-11 | community | – | source |
| Radeon RX 9070 XT | llama 7B | Q4_0 | llama.cpp (Vulkan) | 128.8 | 2055.15 | – | – | tg128 | 2025-08-24 | community | – | source |
| Radeon RX 7900 XTX | llama 7B | Q4_0 | llama.cpp (Vulkan) | 128.79 | 2871.79 | – | – | tg128 | 2025-06-28 | community | – | source |
| Radeon RX 9070 XT | llama 7B | Q4_0 | llama.cpp (Vulkan) | 128.53 | 4046.36 | – | – | tg128 | 2025-10-11 | community | – | source |
| Radeon RX 9070 XT | llama 7B | Q4_0 | llama.cpp (Vulkan) | 128.49 | 3586.22 | – | – | tg128 | 2025-09-23 | community | – | source |
| A100 | llama 7B | Q4_0 | llama.cpp (CUDA) | 128.41 | 1175.43 | – | – | tg128 | 2026-05-26 | community | – | source |
| Radeon RX 9070 XT | llama 7B | Q4_0 | llama.cpp (ROCm) | 128.38 | 5541.12 | – | – | tg128 | 2026-06-05 | community | – | source |
| Radeon RX 7900 XTX | llama 7B | Q4_0 | llama.cpp (ROCm) | 128.25 | 3386.75 | – | – | tg128 | 2025-08-08 | community | – | source |
| GeForce RTX 5070 | llama 7B | Q4_0 | llama.cpp (CUDA) | 128.21 | 5783.44 | – | – | tg128 | 2025-08-20 | community | – | source |
| Radeon RX 9070 XT | llama 7B | Q4_0 | llama.cpp (Vulkan) | 127.69 | 3633.48 | – | – | tg128 | 2025-09-23 | community | – | source |
| GeForce RTX 5070 | llama 7B | Q4_0 | llama.cpp (CUDA) | 127.54 | 5184.75 | – | – | tg128 | 2025-08-20 | community | – | source |
| Radeon RX 7900 XTX | llama 7B | Q4_0 | llama.cpp (ROCm) | 127.43 | 3698.73 | – | – | tg128 | 2025-08-13 | community | – | source |
| Radeon RX 9070 XT | llama 7B | Q4_0 | llama.cpp (Vulkan) | 127.38 | 4124.63 | – | – | tg128 | 2025-10-11 | community | – | source |
| Radeon RX 7900 XTX | llama 7B | Q4_0 | llama.cpp (Vulkan) | 127.32 | 3486.59 | – | – | tg128 | 2025-07-17 | community | – | source |
| Radeon RX 9070 XT | llama 7B | Q4_0 | llama.cpp (ROCm) | 127.26 | 5380.18 | – | – | tg128 | 2026-04-09 | community | – | source |
| Radeon RX 9070 XT | llama 7B | Q4_0 | llama.cpp (Vulkan) | 126.81 | 3493.01 | – | – | tg128 | 2025-12-18 | community | – | source |
| RTX 5070 Ti | llama 7B | Q4_0 | llama.cpp (Vulkan) | 126.67 | 5164.6 | – | – | tg128 | 2025-05-10 | community | – | source |
| Radeon RX 7900 XTX | llama 7B | Q4_0 | llama.cpp (ROCm) | 126.28 | 3730.42 | – | – | tg128 | 2025-12-10 | community | – | source |
| RTX 3090 | llama 7B | Q4_0 | llama.cpp (Vulkan) | 126.06 | 3384.6 | – | – | tg128 | 2025-06-18 | community | – | source |
| Radeon RX 7900 XTX | llama 7B | Q4_0 | llama.cpp (Vulkan) | 126.06 | 1888.77 | – | – | tg128 | 2025-11-20 | community | contradiction | source |
| Radeon RX 7900 XTX | llama 7B | Q4_0 | llama.cpp (Vulkan) | 126.06 | 1888.77 | – | – | tg128 | 2025-11-20 | community | contradiction | source |
| Radeon AI PRO R9700 | llama 7B | Q4_0 | llama.cpp (Vulkan) | 125.96 | 5033.0 | – | – | tg128 | 2026-09-12 | community | – | source |
| RTX 5070 Ti | llama 7B | Q4_0 | llama.cpp (Vulkan) | 125.89 | 6801.97 | – | – | tg128 | 2025-05-12 | community | – | source |
| Radeon RX 7900 XTX | llama 7B | Q4_0 | llama.cpp (ROCm) | 125.84 | 3243.15 | – | – | tg128 | 2025-08-02 | community | – | source |
| Radeon RX 9070 XT | llama 7B | Q4_0 | llama.cpp (Vulkan) | 125.64 | 4945.07 | – | – | tg128 | 2026-03-27 | community | – | source |
| RTX 3090 | llama 7B | Q4_0 | llama.cpp (Vulkan) | 125.6 | 3336.74 | – | – | tg128 | 2025-06-18 | community | – | source |
| Radeon RX 9070 XT | llama 7B | Q4_0 | llama.cpp (Vulkan) | 125.45 | 3530.68 | – | – | tg128 | 2025-08-07 | community | – | source |
| Radeon RX 7800 XT | llama 7B | Q4_0 | llama.cpp (Vulkan) | 124.86 | 2197.05 | – | – | tg128 | 2026-01-29 | community | – | source |
| Radeon RX 9070 XT | llama 7B | Q4_0 | llama.cpp (Vulkan) | 124.75 | 3787.28 | – | – | tg128 | 2025-06-29 | community | – | source |
| Radeon RX 7900 XTX | llama 7B | Q4_0 | llama.cpp (ROCm) | 124.61 | 3674.25 | – | – | tg128 | 2025-08-08 | community | – | source |
| Radeon RX 9070 XT | llama 7B | Q4_0 | llama.cpp (Vulkan) | 124.33 | 3558.71 | – | – | tg128 | 2025-08-07 | community | – | source |
| Radeon RX 7800 XT | llama 7B | Q4_0 | llama.cpp (Vulkan) | 123.89 | 994.51 | – | – | tg128 | 2025-09-27 | community | – | source |
| RTX 3090 | llama 7B | Q4_0 | llama.cpp (Vulkan) | 123.72 | 3301.47 | – | – | tg128 | 2025-01-08 | community | – | source |
| Radeon RX 9070 XT | llama 7B | Q4_0 | llama.cpp (Vulkan) | 123.7 | 2307.03 | – | – | tg128 | 2025-06-23 | community | – | source |
| Radeon RX 7900 XTX | llama 7B | Q4_0 | llama.cpp (Vulkan) | 123.7 | 1941.47 | – | – | tg128 | 2025-11-15 | community | contradiction | source |
| Radeon RX 7900 XTX | llama 7B | Q4_0 | llama.cpp (ROCm) | 123.69 | 3405.36 | – | – | tg128 | 2025-07-27 | community | – | source |
| Radeon RX 7900 XT | llama 7B | Q4_0 | llama.cpp (Vulkan) | 123.18 | 2941.58 | – | – | tg128 | 2025-06-04 | community | – | source |
| Radeon RX 7900 XTX | llama 7B | Q4_0 | llama.cpp (Vulkan) | 122.9 | 2530.61 | – | – | tg128 | 2025-11-20 | community | contradiction | source |
| Radeon RX 7900 XT | llama 7B | Q4_0 | llama.cpp (ROCm) | 122.86 | 3219.16 | – | – | tg128 | 2025-06-04 | community | – | source |
| Radeon RX 7900 XTX | llama 7B | Q4_0 | llama.cpp (ROCm) | 122.71 | 3557.68 | – | – | tg128 | 2025-08-02 | community | – | source |
| RTX 3090 | llama 7B | Q4_0 | llama.cpp (Vulkan) | 122.61 | 4241.77 | – | – | tg128 | 2025-04-12 | community | contradiction | source |
| Radeon RX 7900 XTX | llama 7B | Q4_0 | llama.cpp (ROCm) | 122.56 | 3169.7 | – | – | tg128 | 2025-06-28 | community | – | source |
| Radeon RX 9070 XT | llama 7B | Q4_0 | llama.cpp (Vulkan) | 122.54 | 3399.14 | – | – | tg128 | 2025-06-29 | community | – | source |
| Radeon RX 7800 XT | llama 7B | Q4_0 | llama.cpp (Vulkan) | 122.52 | 1002.13 | – | – | tg128 | 2025-09-27 | community | – | source |
| Radeon RX 7900 XTX | llama 7B | Q4_0 | llama.cpp (ROCm) | 122.06 | 3369.65 | – | – | tg128 | 2025-08-02 | community | – | source |
| Radeon RX 7900 XTX | llama 7B | Q4_0 | llama.cpp (Vulkan) | 122.02 | 2272.76 | – | – | tg128 | 2025-12-11 | community | – | source |
| Radeon RX 7900 XTX | llama 7B | Q4_0 | llama.cpp (ROCm) | 121.87 | 3236.05 | – | – | tg128 | 2025-06-23 | community | – | source |
| A100 | llama 7B | Q4_0 | llama.cpp (Vulkan) | 121.83 | 3103.32 | – | – | tg128 | 2025-05-22 | community | – | source |
| Radeon RX 7900 XTX | llama 7B | Q4_0 | llama.cpp (Vulkan) | 121.62 | 2273.79 | – | – | tg128 | 2025-12-11 | community | – | source |
| Radeon AI PRO R9700 | llama 7B | Q4_0 | llama.cpp (ROCm) | 121.62 | 5039.78 | – | – | tg128 | 2026-09-12 | community | – | source |
| Radeon RX 7900 XTX | llama 7B | Q4_0 | llama.cpp (Vulkan) | 121.61 | 2524.02 | – | – | tg128 | 2025-12-11 | community | – | source |
| Radeon RX 7800 XT | llama 7B | Q4_0 | llama.cpp (Vulkan) | 121.58 | 1638.62 | – | – | tg128 | 2025-09-27 | community | – | source |
| Radeon RX 7900 XTX | llama 7B | Q4_0 | llama.cpp (Vulkan) | 121.57 | 2078.19 | – | – | tg128 | 2025-12-11 | community | contradiction | source |
| Radeon RX 7900 XTX | llama 7B | Q4_0 | llama.cpp (Vulkan) | 121.56 | 2276.59 | – | – | tg128 | 2025-12-11 | community | – | source |
| Radeon RX 7900 XTX | llama 7B | Q4_0 | llama.cpp (Vulkan) | 121.55 | 1869.34 | – | – | tg128 | 2025-12-11 | community | contradiction | source |
| Radeon RX 7900 XTX | llama 7B | Q4_0 | llama.cpp (Vulkan) | 121.5 | 2512.95 | – | – | tg128 | 2025-12-11 | community | – | source |
| Radeon RX 7800 XT | llama 7B | Q4_0 | llama.cpp (Vulkan) | 121.4 | 1595.43 | – | – | tg128 | 2025-09-27 | community | – | source |
| Radeon RX 7800 XT | llama 7B | Q4_0 | llama.cpp (Vulkan) | 121.38 | 1998.36 | – | – | tg128 | 2025-10-08 | community | – | source |
| Radeon RX 7900 XTX | llama 7B | Q4_0 | llama.cpp (Vulkan) | 121.33 | 2521.5 | – | – | tg128 | 2025-12-11 | community | – | source |
| Radeon AI PRO R9700 | llama 7B | Q4_0 | llama.cpp (ROCm) | 121.29 | 2755.01 | – | – | tg128 | 2026-09-12 | community | – | source |
| Radeon RX 9070 XT | llama 7B | Q4_0 | llama.cpp (Vulkan) | 121.25 | 2313.09 | – | – | tg128 | 2025-06-23 | community | – | source |
| Radeon RX 7900 XTX | llama 7B | Q4_0 | llama.cpp (ROCm) | 121.14 | 3401.14 | – | – | tg128 | 2025-07-27 | community | – | source |
| Radeon RX 9070 XT | llama 7B | Q4_0 | llama.cpp (ROCm) | 121.04 | 4306.74 | – | – | tg128 | 2026-04-09 | community | – | source |
| Radeon RX 7900 XTX | llama 7B | Q4_0 | llama.cpp (ROCm) | 120.88 | 3379.59 | – | – | tg128 | 2025-07-27 | community | – | source |
| Radeon RX 7900 XTX | llama 7B | Q4_0 | llama.cpp (ROCm) | 120.73 | 3638.45 | – | – | tg128 | 2025-07-27 | community | – | source |
| Radeon RX 7900 XT | llama 7B | Q4_0 | llama.cpp (Vulkan) | 120.62 | 2701.13 | – | – | tg128 | 2025-06-04 | community | – | source |
| A100 | llama 7B | Q4_0 | llama.cpp (Vulkan) | 120.53 | 3164.55 | – | – | tg128 | 2025-05-22 | community | – | source |
| Radeon RX 9070 XT | llama 7B | Q4_0 | llama.cpp (Vulkan) | 120.52 | 1707.58 | – | – | tg128 | 2025-06-29 | community | – | source |
| Radeon RX 9070 XT | llama 7B | Q4_0 | llama.cpp (Vulkan) | 120.46 | 2296.83 | – | – | tg128 | 2025-05-26 | community | – | source |
| RTX 3090 | llama 7B | Q4_0 | llama.cpp (Vulkan) | 120.44 | 4516.92 | – | – | tg128 | 2025-04-12 | community | contradiction | source |
| Radeon RX 7800 XT | llama 7B | Q4_0 | llama.cpp (Vulkan) | 120.25 | 1573.21 | – | – | tg128 | 2025-09-27 | community | – | source |
| Radeon RX 7800 XT | llama 7B | Q4_0 | llama.cpp (Vulkan) | 120.12 | 1659.94 | – | – | tg128 | 2025-09-27 | community | – | source |
| Radeon RX 7900 XTX | llama 7B | Q4_0 | llama.cpp (Vulkan) | 119.98 | 1873.97 | – | – | tg128 | 2025-11-20 | community | contradiction | source |
| Radeon RX 7900 XTX | llama 7B | Q4_0 | llama.cpp (Vulkan) | 119.98 | 1873.97 | – | – | tg128 | 2025-11-20 | community | contradiction | source |
| Radeon RX 7900 XTX | llama 7B | Q4_0 | llama.cpp (ROCm) | 119.97 | 3408.06 | – | – | tg128 | 2025-11-20 | community | contradiction | source |
| Radeon RX 9070 XT | llama 7B | Q4_0 | llama.cpp (ROCm) | 119.97 | 4824.33 | – | – | tg128 | 2026-03-25 | community | – | source |
| M5 Max (40 GPU) | llama 7B | Q4_0 | llama.cpp (Metal) | 119.92 | 3219.99 | – | – | – | 2023-11-22 | community | – | source |
| M5 Max | llama 7B | Q4_0 | llama.cpp (Metal) | 119.92 | 3219.99 | – | – | tg128 | 2026-08-25 | community | – | source |
| Radeon RX 7800 XT | llama 7B | Q4_0 | llama.cpp (Vulkan) | 119.89 | 1919.21 | – | – | tg128 | 2025-10-08 | community | – | source |
| Radeon RX 9070 | llama 7B | Q4_0 | llama.cpp (Vulkan) | 119.71 | 3164.1 | – | – | tg128 | 2025-08-15 | community | – | source |
| Radeon RX 7900 XTX | llama 7B | Q4_0 | llama.cpp (Vulkan) | 119.69 | 1892.58 | – | – | tg128 | 2025-11-15 | community | contradiction | source |
| Radeon RX 9070 | llama 7B | Q4_0 | llama.cpp (Vulkan) | 119.51 | 2859.98 | – | – | tg128 | 2025-08-15 | community | – | source |
| M5 Max | llama 7B | Q4_0 | llama.cpp (Metal) | 119.08 | 3347.44 | – | – | tg128 | 2026-06-22 | community | – | source |
| M5 Ultra | llama 7B 6.74B | Q8_0 | llama.cpp (Metal) | 119.01 | 4910.55 | – | – | tg128 | 2026-09-30 | community | – | source |
| Radeon RX 7900 XTX | llama 7B | Q4_0 | llama.cpp (ROCm) | 119.0 | 3429.68 | – | – | tg128 | 2025-12-10 | community | – | source |
| Radeon RX 7900 XTX | llama 7B | Q4_0 | llama.cpp (ROCm) | 118.71 | 3573.3 | – | – | tg128 | 2025-08-02 | community | – | source |
| RTX 3090 | llama 7B | Q4_0 | llama.cpp (Vulkan) | 118.5 | 4368.74 | – | – | tg128 | 2025-01-16 | community | – | source |
| Radeon RX 7800 XT | llama 7B | Q4_0 | llama.cpp (Vulkan) | 118.27 | 2017.33 | – | – | tg128 | 2026-01-29 | community | – | source |
| Radeon RX 7900 XTX | llama 7B | Q4_0 | llama.cpp (ROCm) | 118.23 | 3323.43 | – | – | tg128 | 2025-06-28 | community | – | source |
| Radeon RX 9070 XT | llama 7B | Q4_0 | llama.cpp (Vulkan) | 117.72 | 2336.07 | – | – | tg128 | 2025-05-26 | community | – | source |
| Radeon RX 7900 XT | llama 7B | Q4_0 | llama.cpp (Vulkan) | 117.65 | 2609.51 | – | – | tg128 | 2025-07-30 | community | – | source |
| Radeon RX 7900 XTX | llama 7B | Q4_0 | llama.cpp (ROCm) | 117.61 | 3409.8 | – | – | tg128 | 2025-06-23 | community | – | source |
| Radeon RX 7900 XT | llama 7B | Q4_0 | llama.cpp (Vulkan) | 117.57 | 2076.96 | – | – | tg128 | 2025-08-23 | community | – | source |
| Radeon RX 7900 XT | llama 7B | Q4_0 | llama.cpp (Vulkan) | 117.45 | 2823.35 | – | – | tg128 | 2025-07-30 | community | – | source |
| Radeon RX 9070 XT | llama 7B | Q4_0 | llama.cpp (Vulkan) | 117.22 | 2206.24 | – | – | tg128 | 2025-03-18 | community | – | source |
| GeForce RTX 5090 | llama 7B | Q4_0 | llama.cpp (CUDA) | 117.16 | 5007.9 | – | – | tg128 | 2026-05-26 | community | contradiction | source |
| Radeon RX 7900 XTX | llama 7B | Q4_0 | llama.cpp (ROCm) | 117.12 | 3605.23 | – | – | tg128 | 2025-07-27 | community | – | source |
| Radeon RX 7900 XT | llama 7B | Q4_0 | llama.cpp (ROCm) | 117.11 | 3267.35 | – | – | tg128 | 2025-06-04 | community | – | source |
| Radeon RX 7800 XT | llama 7B | Q4_0 | llama.cpp (Vulkan) | 116.97 | 2053.11 | – | – | tg128 | 2025-09-27 | community | – | source |
| Radeon RX 9070 XT | llama 7B | Q4_0 | llama.cpp (ROCm) | 116.95 | 4733.95 | – | – | tg128 | 2026-06-05 | community | – | source |
| Radeon RX 7900 XT | llama 7B | Q4_0 | llama.cpp (Vulkan) | 116.94 | 1662.15 | – | – | tg128 | 2025-08-23 | community | – | source |
| Radeon RX 7900 XTX | llama 7B | Q4_0 | llama.cpp (ROCm) | 116.83 | 3447.07 | – | – | tg128 | 2025-11-20 | community | contradiction | source |
| Radeon RX 7900 XTX | llama 7B | Q4_0 | llama.cpp (ROCm) | 116.83 | 3447.07 | – | – | tg128 | 2025-11-20 | community | contradiction | source |
| Radeon RX 9070 XT | llama 7B | Q4_0 | llama.cpp (Vulkan) | 116.82 | 1689.15 | – | – | tg128 | 2025-06-29 | community | – | source |
| Radeon RX 7900 GRE | llama 7B | Q4_0 | llama.cpp (Vulkan) | 116.7 | 2711.43 | – | – | tg128 | 2026-03-18 | community | – | source |
| Arc Pro B70 | llama 7B | Q4_0 | llama.cpp (Vulkan) | 116.67 | 4651.26 | – | – | tg128 | 2026-08-26 | community | – | source |
| Radeon RX 7800 XT | llama 7B | Q4_0 | llama.cpp (Vulkan) | 116.61 | 1501.02 | – | – | tg128 | 2025-09-27 | community | – | source |
| Radeon RX 9070 XT | llama 7B | Q4_0 | llama.cpp (ROCm) | 116.5 | 5167.28 | – | – | tg128 | 2026-03-25 | community | – | source |
| Radeon RX 9070 XT | llama 7B | Q4_0 | llama.cpp (ROCm) | 116.46 | 5166.48 | – | – | tg128 | 2026-04-20 | community | – | source |
| Radeon RX 7900 XT | llama 7B | Q4_0 | llama.cpp (ROCm) | 116.15 | 3098.38 | – | – | tg128 | 2025-07-30 | community | – | source |
| Radeon RX 7900 GRE | llama 7B | Q4_0 | llama.cpp (Vulkan) | 116.11 | 2336.31 | – | – | tg128 | 2025-12-18 | community | – | source |
| Radeon RX 7900 XT | llama 7B | Q4_0 | llama.cpp (ROCm) | 116.0 | 2967.12 | – | – | tg128 | 2025-08-01 | community | – | source |
| Radeon RX 9070 | llama 7B | Q4_0 | llama.cpp (ROCm) | 115.84 | 4074.98 | – | – | tg128 | 2025-10-23 | community | – | source |
| Radeon RX 7800 XT | llama 7B | Q4_0 | llama.cpp (Vulkan) | 115.71 | 1967.78 | – | – | tg128 | 2025-09-27 | community | – | source |
| Radeon RX 7800 XT | llama 7B | Q4_0 | llama.cpp (Vulkan) | 115.55 | 881.05 | – | – | tg128 | 2025-09-27 | community | – | source |
| M3 Ultra | llama 7B | Q4_0 | llama.cpp (Vulkan) | 115.54 | 1116.83 | – | – | tg128 | 2025-04-27 | community | – | source |
| Radeon RX 9070 | llama 7B | Q4_0 | llama.cpp (ROCm) | 115.32 | 2452.68 | – | – | tg128 | 2025-10-23 | community | – | source |
| Radeon RX 9070 | llama 7B | Q4_0 | llama.cpp (ROCm) | 115.11 | 3901.21 | – | – | tg128 | 2025-10-23 | community | – | source |
| M3 Ultra | llama 7B | Q4_0 | llama.cpp (Vulkan) | 114.55 | 1116.48 | – | – | tg128 | 2025-04-27 | community | – | source |
| Radeon RX 9070 | llama 7B | Q4_0 | llama.cpp (ROCm) | 114.48 | 2381.77 | – | – | tg128 | 2025-10-23 | community | – | source |
| Radeon RX 7800 XT | llama 7B | Q4_0 | llama.cpp (Vulkan) | 114.18 | 1258.58 | – | – | tg128 | 2025-05-14 | community | – | source |
| Radeon RX 7800 XT | llama 7B | Q4_0 | llama.cpp (Vulkan) | 114.12 | 1432.7 | – | – | tg128 | 2025-05-14 | community | – | source |
| Radeon RX 7900 XTX | llama 7B | Q4_0 | llama.cpp (ROCm) | 114.02 | 3190.19 | – | – | tg128 | 2025-11-20 | community | contradiction | source |
| M1 Ultra | llama 7B | Q4_0 | llama.cpp (Metal) | 113.79 | 1179.89 | – | – | tg128 | 2026-08-28 | community | – | source |
| RTX 4070 Ti | llama 7B | Q4_0 | llama.cpp (Vulkan) | 113.52 | 5515.26 | – | – | tg128 | 2026-01-17 | community | – | source |
| Radeon RX 7800 XT | llama 7B | Q4_0 | llama.cpp (Vulkan) | 113.29 | 1003.78 | – | – | tg128 | 2025-09-27 | community | – | source |
| Radeon RX 7900 XT | llama 7B | Q4_0 | llama.cpp (ROCm) | 112.75 | 3163.24 | – | – | tg128 | 2025-08-01 | community | – | source |
| Radeon AI PRO R9700 | llama 7B | Q4_0 | llama.cpp (ROCm) | 112.72 | 4665.08 | – | – | tg128 | 2026-09-12 | community | – | source |
| Radeon RX 7900 GRE | llama 7B | Q4_0 | llama.cpp (Vulkan) | 112.59 | 1632.95 | – | – | tg128 | 2025-10-15 | community | – | source |
| Arc Pro B70 | llama 7B | Q4_0 | llama.cpp (Vulkan) | 112.43 | 4271.32 | – | – | tg128 | 2026-08-26 | community | – | source |
| GeForce RTX 5090 | llama 7B | Q4_0 | llama.cpp (CUDA) | 112.36 | 5056.95 | – | – | tg128 | 2026-05-26 | community | contradiction | source |
| Radeon RX 7900 XT | llama 7B | Q4_0 | llama.cpp (ROCm) | 112.3 | 3261.75 | – | – | tg128 | 2025-07-30 | community | – | source |
| Radeon RX 7900 XTX | llama 7B | Q4_0 | llama.cpp (ROCm) | 112.18 | 2817.81 | – | – | tg128 | 2025-09-17 | community | – | source |
| Arc Pro B70 | llama 7B | Q4_0 | llama.cpp (SYCL) | 112.08 | 1434.13 | – | – | tg128 | 2026-08-26 | community | – | source |
| Radeon RX 7800 XT | llama 7B | Q4_0 | llama.cpp (Vulkan) | 112.01 | 1244.85 | – | – | tg128 | 2025-05-14 | community | – | source |
| RTX 4070 SUPER | llama 7B | Q4_0 | llama.cpp (Vulkan) | 112.0 | 5198.91 | – | – | tg128 | 2026-01-16 | community | – | source |
| Radeon RX 7800 XT | llama 7B | Q4_0 | llama.cpp (Vulkan) | 111.97 | 1531.85 | – | – | tg128 | 2025-05-14 | community | – | source |
| Radeon RX 7800 XT | llama 7B | Q4_0 | llama.cpp (Vulkan) | 111.97 | 728.81 | – | – | tg128 | 2025-09-27 | community | – | source |
| Radeon RX 7900 GRE | llama 7B | Q4_0 | llama.cpp (Vulkan) | 111.32 | 1420.2 | – | – | tg128 | 2025-09-11 | community | – | source |
| Radeon RX 7900 GRE | llama 7B | Q4_0 | llama.cpp (Vulkan) | 111.31 | 1604.81 | – | – | tg128 | 2025-10-15 | community | – | source |
| Arc Pro B70 | llama 7B | Q4_0 | llama.cpp (SYCL) | 111.3 | 1133.3 | – | – | tg128 | 2026-08-26 | community | – | source |
| Arc Pro B70 | llama 7B | Q4_0 | llama.cpp (SYCL) | 111.13 | 3271.16 | – | – | tg128 | 2026-09-16 | community | – | source |
| Arc Pro B70 | llama 7B | Q4_0 | llama.cpp (SYCL) | 111.13 | 3271.16 | – | – | tg128 | – | community | – | source |
| Radeon RX 7900 GRE | llama 7B | Q4_0 | llama.cpp (Vulkan) | 110.91 | 2504.25 | – | – | tg128 | 2026-03-18 | community | – | source |
| Radeon RX 7900 GRE | llama 7B | Q4_0 | llama.cpp (Vulkan) | 110.8 | 1760.15 | – | – | tg128 | 2025-12-14 | community | – | source |
| Radeon RX 7900 GRE | llama 7B | Q4_0 | llama.cpp (Vulkan) | 110.73 | 2372.49 | – | – | tg128 | 2025-12-14 | community | – | source |
| Radeon RX 7900 XTX | llama 7B | Q4_0 | llama.cpp (ROCm) | 110.7 | 3053.77 | – | – | tg128 | 2025-09-17 | community | – | source |
| RTX 4070 Ti | llama 7B | Q4_0 | llama.cpp (Vulkan) | 110.53 | 4981.44 | 221 | – | tg128 | 2026-01-17 | community | – | source |
| Radeon RX 7900 GRE | llama 7B | Q4_0 | llama.cpp (Vulkan) | 110.5 | 2251.37 | – | – | tg128 | 2025-12-18 | community | – | source |
| Arc Pro B70 | llama 7B | Q4_0 | llama.cpp (SYCL) | 109.99 | 1135.88 | – | – | tg128 | 2026-08-26 | community | – | source |
| Radeon RX 7900 XTX | llama 7B | Q4_0 | llama.cpp (ROCm) | 109.92 | 3226.16 | – | – | tg128 | 2025-11-20 | community | contradiction | source |
| Radeon RX 7900 XTX | llama 7B | Q4_0 | llama.cpp (ROCm) | 109.92 | 3226.16 | – | – | tg128 | 2025-11-20 | community | contradiction | source |
| Radeon RX 9070 XT | llama 7B | Q4_0 | llama.cpp (ROCm) | 109.81 | 4661.14 | – | – | tg128 | 2026-03-25 | community | – | source |
| Arc Pro B70 | llama 7B | Q4_0 | llama.cpp (SYCL) | 109.72 | 1382.27 | – | – | tg128 | 2026-08-26 | community | – | source |
| V100 | llama 7B | Q4_0 | llama.cpp (CUDA) | 109.49 | 380.4 | – | – | tg128 | 2026-05-26 | community | – | source |
| V100 | llama 7B | Q4_0 | llama.cpp (CUDA) | 109.46 | 1156.76 | – | – | tg128 | 2026-05-26 | community | – | source |
| M2 | llama 7B | Q4_0 | llama.cpp (Vulkan) | 109.21 | 792.74 | – | – | tg128 | 2025-10-18 | community | – | source |
| RTX 4070 SUPER | llama 7B | Q4_0 | llama.cpp (Vulkan) | 108.74 | 4608.2 | – | – | tg128 | 2026-01-16 | community | – | source |
| RTX 3090 | llama 7B | Q4_0 | llama.cpp (Vulkan) | 108.45 | 783.94 | – | – | tg128 | 2025-09-01 | community | – | source |
| Radeon RX 9070 XT | llama 7B | Q4_0 | llama.cpp (ROCm) | 108.45 | 4766.48 | – | – | tg128 | 2026-04-01 | community | – | source |
| A100 | llama 7B | Q4_0 | llama.cpp (Vulkan) | 108.39 | 4040.77 | – | – | tg128 | 2026-08-14 | community | – | source |
| Radeon RX 6900 XT | llama 7B | Q4_0 | llama.cpp (Vulkan) | 108.0 | 1901.2 | – | – | tg128 | 2025-09-09 | community | – | source |
| Radeon RX 6900 XT | llama 7B | Q4_0 | llama.cpp (Vulkan) | 108.0 | 1901.2 | – | – | tg128 | 2025-09-09 | community | – | source |
| Radeon RX 7800 XT | llama 7B | Q4_0 | llama.cpp (Vulkan) | 107.53 | 1260.54 | – | – | tg128 | 2025-02-21 | community | – | source |
| RTX 3090 | llama 7B | Q4_0 | llama.cpp (Vulkan) | 107.32 | 4761.91 | – | – | tg128 | 2025-01-16 | community | – | source |
| Radeon RX 7900 XTX | llama 7B | Q4_0 | llama.cpp (ROCm) | 107.16 | 1934.05 | – | – | tg128 | 2025-07-27 | community | contradiction | source |
| Arc Pro B70 | llama 7B | Q4_0 | llama.cpp (SYCL) | 107.11 | 1324.76 | – | – | tg128 | 2026-09-16 | community | – | source |
| Arc Pro B70 | llama 7B | Q4_0 | llama.cpp (SYCL) | 107.11 | 1324.76 | – | – | tg128 | – | community | – | source |
| Radeon RX 7900 XTX | llama 7B | Q4_0 | llama.cpp (ROCm) | 107.02 | 3044.46 | – | – | tg128 | 2025-05-20 | community | – | source |
| Arc Pro B70 | llama 7B | Q4_0 | llama.cpp (SYCL) | 106.99 | 3009.58 | – | – | tg128 | 2026-08-26 | community | – | source |
| Radeon RX 9070 XT | llama 7B | Q4_0 | llama.cpp (ROCm) | 106.7 | 4753.07 | – | – | tg128 | 2026-03-25 | community | – | source |
| Arc Pro B70 | llama 7B | Q4_0 | llama.cpp (SYCL) | 106.69 | 971.17 | – | – | tg128 | 2026-08-26 | community | – | source |
| Radeon RX 9070 XT | llama 7B | Q4_0 | llama.cpp (ROCm) | 106.63 | 4732.96 | – | – | tg128 | 2026-04-20 | community | – | source |
| RX 6800 | llama 7B | Q4_0 | llama.cpp (Vulkan) | 106.49 | 1597.95 | – | – | tg128 | 2026-10-01 | community | – | source |
| Radeon RX 6900 XT | llama 7B | Q4_0 | llama.cpp (Vulkan) | 106.15 | 1761.93 | – | – | tg128 | 2025-09-09 | community | – | source |
| Radeon RX 6900 XT | llama 7B | Q4_0 | llama.cpp (Vulkan) | 106.15 | 1761.93 | – | – | tg128 | 2025-09-09 | community | – | source |
| RTX 3090 | llama 7B | Q4_0 | llama.cpp (Vulkan) | 106.03 | 912.47 | – | – | tg128 | 2025-09-01 | community | – | source |
| Radeon RX 7900 GRE | llama 7B | Q4_0 | llama.cpp (Vulkan) | 105.91 | 2411.47 | – | – | tg128 | 2025-12-14 | community | – | source |
| Arc Pro B70 | llama 7B | Q4_0 | llama.cpp (SYCL) | 105.64 | 2830.14 | – | – | tg128 | 2026-08-26 | community | – | source |
| Arc Pro B70 | llama 7B | Q4_0 | llama.cpp (SYCL) | 105.45 | 970.74 | – | – | tg128 | 2026-08-26 | community | – | source |
| M2 | llama 7B | Q4_0 | llama.cpp (Vulkan) | 105.06 | 796.29 | – | – | tg128 | 2025-10-18 | community | – | source |
| Radeon RX 7900 GRE | llama 7B | Q4_0 | llama.cpp (Vulkan) | 104.75 | 1711.33 | – | – | tg128 | 2025-12-14 | community | – | source |
| RTX 5060 Ti | llama 7B | Q4_0 | llama.cpp (Vulkan) | 104.42 | 1907.81 | – | – | tg128 | 2025-11-04 | community | – | source |
| Quadro RTX 6000 | llama 7B | Q4_0 | llama.cpp (CUDA) | 104.38 | 3053.96 | – | – | tg128 | 2025-09-15 | community | – | source |
| Radeon RX 7800 XT | llama 7B | Q4_0 | llama.cpp (Vulkan) | 103.98 | 883.2 | – | – | tg128 | 2025-02-21 | community | – | source |
| A4500 | llama 7B | Q4_0 | llama.cpp (CUDA) | 103.0 | 3453.1 | – | – | tg128 | 2025-08-14 | community | – | source |
| Quadro RTX 6000 | llama 7B | Q4_0 | llama.cpp (CUDA) | 102.77 | 2751.18 | – | – | tg128 | 2025-09-15 | community | – | source |
| RTX 5060 Ti | llama 7B | Q4_0 | llama.cpp (Vulkan) | 102.66 | 1986.47 | – | – | tg128 | 2025-11-04 | community | – | source |
| Radeon RX 9070 XT | llama 7B | Q4_0 | llama.cpp (ROCm) | 102.55 | 4336.99 | – | – | tg128 | 2025-11-15 | community | – | source |
| A100 | llama 7B | Q4_0 | llama.cpp (Vulkan) | 102.46 | 3222.01 | – | – | tg128 | 2026-08-14 | community | – | source |
| Radeon RX 9070 XT | llama 7B | Q4_0 | llama.cpp (ROCm) | 102.44 | 5009.54 | – | – | tg128 | 2026-03-06 | community | – | source |
| Radeon RX 7900 GRE | llama 7B | Q4_0 | llama.cpp (ROCm) | 102.38 | 2982.56 | – | – | tg128 | 2026-03-18 | community | – | source |
| RTX 5060 Ti | llama 7B | Q4_0 | llama.cpp (Vulkan) | 102.31 | 1741.01 | – | – | tg128 | 2025-11-04 | community | – | source |
| M3 Ultra | llama 7B | Q4_0 | llama.cpp (Metal) | 102.28 | 1574.09 | – | – | tg128 | 2025-04-27 | community | – | source |
| Radeon RX 7900 GRE | llama 7B | Q4_0 | llama.cpp (Vulkan) | 101.73 | 1465.45 | – | – | tg128 | 2025-09-11 | community | – | source |
| Radeon RX 9070 XT | llama 7B | Q4_0 | llama.cpp (ROCm) | 101.61 | 5077.32 | – | – | tg128 | 2025-12-22 | community | – | source |
| RX 6800 | llama 7B | Q4_0 | llama.cpp (Vulkan) | 101.47 | 1593.97 | – | – | tg128 | 2026-10-01 | community | – | source |
| Radeon RX 6900 XT | llama 7B | Q4_0 | llama.cpp (Vulkan) | 101.42 | 1257.98 | – | – | tg128 | 2025-01-18 | community | – | source |
| RTX 5060 Ti | llama 7B | Q4_0 | llama.cpp (Vulkan) | 101.35 | 1678.98 | – | – | tg128 | 2025-11-04 | community | – | source |
| Radeon RX 9070 XT | llama 7B | Q4_0 | llama.cpp (ROCm) | 101.27 | 5055.19 | – | – | tg128 | 2025-11-26 | community | – | source |
| RTX 2080 Ti | llama 7B | Q4_0 | llama.cpp (Vulkan) | 100.99 | 1936.25 | – | – | tg128 | 2025-08-27 | community | – | source |
| Radeon RX 7800 XT | llama 7B | Q4_0 | llama.cpp (ROCm) | 100.94 | 2151.81 | – | – | tg128 | 2025-08-01 | community | – | source |
| Radeon RX 7800 XT | llama 7B | Q4_0 | llama.cpp (ROCm) | 100.94 | 2151.81 | – | – | tg128 | 2025-08-01 | community | – | source |
| Radeon RX 7900 XTX | llama 7B | Q4_0 | llama.cpp (Vulkan) | 100.62 | 1668.19 | – | – | tg128 | 2025-01-08 | community | contradiction | source |
| GeForce RTX 5060 | llama 7B | Q4_0 | llama.cpp (CUDA) | 100.61 | 3799.47 | – | – | tg128 | 2026-06-19 | community | – | source |
| Radeon RX 6800 XT | llama 7B | Q4_0 | llama.cpp (Vulkan) | 100.5 | 1704.79 | – | – | tg128 | 2025-09-01 | community | – | source |
| Radeon RX 6800 XT | llama 7B | Q4_0 | llama.cpp (Vulkan) | 100.32 | 1752.92 | – | – | tg128 | 2025-09-01 | community | – | source |
| Radeon RX 7900 XTX | llama 7B | Q4_0 | llama.cpp (ROCm) | 99.96 | 1601.25 | – | – | tg128 | 2026-03-01 | community | – | source |
| V100 | llama 7B | Q4_0 | llama.cpp (CUDA) | 99.95 | 977.33 | – | – | tg128 | 2026-05-26 | community | – | source |
| V100 | llama 7B | Q4_0 | llama.cpp (CUDA) | 99.84 | 354.07 | – | – | tg128 | 2026-05-26 | community | – | source |
| Radeon RX 7800 XT | llama 7B | Q4_0 | llama.cpp (ROCm) | 99.55 | 2109.38 | – | – | tg128 | 2025-08-08 | community | – | source |
| Radeon RX 9070 | llama 7B | Q4_0 | llama.cpp (ROCm) | 99.39 | 2361.1 | – | – | tg128 | 2025-08-17 | community | – | source |
| Radeon RX 6900 XT | llama 7B | Q4_0 | llama.cpp (Vulkan) | 98.89 | 1121.36 | – | – | tg128 | 2025-02-05 | community | – | source |
| V620 | llama 7B | Q4_0 | llama.cpp (ROCm) | 98.78 | 2173.13 | – | – | tg128 | 2026-09-10 | community | – | source |
| Radeon RX 7900 XTX | llama 7B | Q4_0 | llama.cpp (ROCm) | 98.71 | 5025.17 | – | – | tg128 | 2026-03-01 | community | – | source |
| A4500 | llama 7B | Q4_0 | llama.cpp (CUDA) | 98.45 | 2960.88 | – | – | tg128 | 2025-08-14 | community | – | source |
| Radeon RX 6900 XT | llama 7B | Q4_0 | llama.cpp (Metal) | 98.27 | 326.29 | – | – | tg128 | 2025-08-11 | community | – | source |
| Radeon RX 7800 XT | llama 7B | Q4_0 | llama.cpp (Vulkan) | 98.15 | 2091.34 | – | – | tg128 | 2025-05-14 | community | contradiction | source |
| Radeon RX 7800 XT | llama 7B | Q4_0 | llama.cpp (Vulkan) | 98.13 | 1487.1 | – | – | tg128 | 2025-05-14 | community | contradiction | source |
| Radeon RX 9070 XT | llama 7B | Q4_0 | llama.cpp (ROCm) | 98.13 | 4244.92 | – | – | tg128 | 2025-11-15 | community | – | source |
| Radeon RX 6800 | llama 7B | Q4_0 | llama.cpp (Vulkan) | 98.11 | 949.45 | – | – | tg128 | 2026-02-12 | community | – | source |
| Radeon RX 9070 XT | llama 7B | Q4_0 | llama.cpp (ROCm) | 98.11 | 3840.62 | – | – | tg128 | 2026-03-06 | community | – | source |
| Radeon 8060S | llama 7B | Q4_0 | llama.cpp (ROCm) | 97.98 | 4773.07 | – | – | tg128 | 2025-11-26 | community | – | source |
| Radeon RX 7900 XTX | llama 7B | Q4_0 | llama.cpp (Vulkan) | 97.96 | 1566.38 | – | – | tg128 | 2025-01-08 | community | contradiction | source |
| RTX 2080 Ti | llama 7B | Q4_0 | llama.cpp (Vulkan) | 97.58 | 1888.24 | – | – | tg128 | 2025-08-27 | community | – | source |
| Radeon RX 7900 GRE | llama 7B | Q4_0 | llama.cpp (ROCm) | 97.53 | 1598.79 | – | – | tg128 | 2025-10-15 | community | – | source |
| A4500 | llama 7B | Q4_0 | llama.cpp (CUDA) | 97.32 | 2827.2 | – | – | tg128 | 2025-08-14 | community | – | source |
| GeForce RTX 5060 Ti | llama 7B | Q4_0 | llama.cpp (CUDA) | 97.32 | 4613.86 | – | – | tg128 | 2026-03-06 | community | – | source |
| Radeon RX 6800 | llama 7B | Q4_0 | llama.cpp (Vulkan) | 97.3 | 933.71 | – | – | tg128 | 2026-02-12 | community | – | source |
| Radeon RX 9070 XT | llama 7B | Q4_0 | llama.cpp (ROCm) | 97.28 | 4903.51 | – | – | tg128 | 2025-11-26 | community | – | source |
| Radeon RX 9070 XT | llama 7B | Q4_0 | llama.cpp (ROCm) | 97.27 | 5025.74 | – | – | tg128 | 2026-03-05 | community | – | source |
| Radeon RX 9070 XT | llama 7B | Q4_0 | llama.cpp (ROCm) | 97.21 | 4944.22 | – | – | tg128 | 2025-12-22 | community | – | source |
| Radeon RX 9070 | llama 7B | Q4_0 | llama.cpp (ROCm) | 97.1 | 1147.66 | – | – | tg128 | 2025-08-17 | community | – | source |
| Radeon RX 7800 XT | llama 7B | Q4_0 | llama.cpp (Vulkan) | 97.07 | 2026.84 | – | – | tg128 | 2025-08-08 | community | – | source |
| Radeon RX 9060 XT | llama 7B | Q4_0 | llama.cpp (Vulkan) | 97.01 | 3912.65 | – | – | tg128 | 2026-02-03 | community | – | source |
| Radeon RX 7800 XT | llama 7B | Q4_0 | llama.cpp (Vulkan) | 96.89 | 2145.6 | – | – | tg128 | 2025-08-01 | community | – | source |
| Radeon RX 7800 XT | llama 7B | Q4_0 | llama.cpp (Vulkan) | 96.89 | 2145.6 | – | – | tg128 | 2025-08-01 | community | – | source |
| Radeon RX 7800 XT | llama 7B | Q4_0 | llama.cpp (ROCm) | 96.8 | 2296.72 | – | – | tg128 | 2025-08-08 | community | – | source |
| GeForce RTX 5060 | llama 7B | Q4_0 | llama.cpp (CUDA) | 96.7 | 3269.08 | – | – | tg128 | 2026-06-19 | community | – | source |
| Radeon RX 6800 | llama 7B | Q4_0 | llama.cpp (Vulkan) | 96.65 | 1749.46 | – | – | tg128 | 2026-02-13 | community | – | source |
| M4 Max | llama 7B | Q4_0 | llama.cpp (Metal) | 96.62 | 823.99 | – | – | tg128 | 2025-07-15 | community | – | source |
| Radeon RX 7800 XT | llama 7B | Q4_0 | llama.cpp (Vulkan) | 96.61 | 2055.58 | – | – | tg128 | 2025-08-08 | community | – | source |
| Radeon RX 7900 GRE | llama 7B | Q4_0 | llama.cpp (ROCm) | 96.53 | 2708.46 | – | – | tg128 | 2026-03-18 | community | – | source |
| GeForce RTX 3060 Ti | llama 7B | Q4_0 | llama.cpp (CUDA) | 96.5 | 2990.99 | – | – | tg128 | 2026-09-19 | community | – | source |
| Radeon RX 7900 GRE | llama 7B | Q4_0 | llama.cpp (ROCm) | 96.07 | 1456.98 | – | – | tg128 | 2025-10-15 | community | – | source |
| Radeon RX 7800 XT | llama 7B | Q4_0 | llama.cpp (Vulkan) | 96.03 | 2063.66 | – | – | tg128 | 2025-08-01 | community | – | source |
| Radeon RX 7800 XT | llama 7B | Q4_0 | llama.cpp (Vulkan) | 96.03 | 2063.66 | – | – | tg128 | 2025-08-01 | community | – | source |
| Radeon RX 7800 XT | llama 7B | Q4_0 | llama.cpp (ROCm) | 95.99 | 2304.63 | – | – | tg128 | 2025-08-01 | community | – | source |
| Radeon RX 7800 XT | llama 7B | Q4_0 | llama.cpp (ROCm) | 95.99 | 2304.63 | – | – | tg128 | 2025-08-01 | community | – | source |
| RTX 5060 Ti | llama 7B | Q4_0 | llama.cpp (Vulkan) | 95.98 | 3848.92 | – | – | tg128 | 2026-02-03 | community | – | source |
| A4500 | llama 7B | Q4_0 | llama.cpp (CUDA) | 95.87 | 3382.11 | – | – | tg128 | 2025-08-14 | community | – | source |
| GeForce RTX 5060 Ti | llama 7B | Q4_0 | llama.cpp (Vulkan) | 95.84 | 3731.36 | – | – | tg128 | 2026-05-08 | community | – | source |
| Radeon RX 6900 XT | llama 7B | Q4_0 | llama.cpp (Vulkan) | 95.7 | 1018.94 | – | – | tg128 | 2025-01-18 | community | – | source |
| GeForce RTX 5060 Ti | llama 7B | Q4_0 | llama.cpp (CUDA) | 95.69 | 4427.42 | – | – | tg128 | 2026-05-08 | community | – | source |
| Radeon RX 6800 | llama 7B | Q4_0 | llama.cpp (Vulkan) | 95.61 | 1698.69 | – | – | tg128 | 2026-02-13 | community | – | source |
| Radeon RX 6800 XT | llama 7B | Q4_0 | llama.cpp (Vulkan) | 95.56 | 1533.6 | – | – | tg128 | 2025-04-12 | community | – | source |
| Radeon RX 7800 XT | llama 7B | Q4_0 | llama.cpp (Vulkan) | 95.2 | 949.34 | – | – | tg128 | 2025-05-14 | community | contradiction | source |
| M4 Max | llama 7B | Q4_0 | llama.cpp (Metal) | 95.19 | 949.34 | – | – | tg128 | 2025-07-15 | community | – | source |
| Radeon RX 7800 XT | llama 7B | Q4_0 | llama.cpp (Vulkan) | 95.17 | 1955.91 | – | – | tg128 | 2025-05-14 | community | contradiction | source |
| GeForce RTX 5060 Ti | llama 7B | Q4_0 | llama.cpp (CUDA) | 95.1 | 4430.01 | – | – | tg128 | 2026-05-08 | community | – | source |
| Radeon RX 7900 XTX | llama 7B | Q4_0 | llama.cpp (ROCm) | 95.04 | 4932.17 | – | – | tg128 | 2026-03-01 | community | – | source |
| Radeon RX 6800 XT | llama 7B | Q4_0 | llama.cpp (Vulkan) | 94.81 | 625.53 | – | – | tg128 | 2025-05-07 | community | – | source |
| Radeon RX 9070 XT | llama 7B | Q4_0 | llama.cpp (ROCm) | 94.32 | 3845.97 | – | – | tg128 | 2026-03-05 | community | – | source |
| RTX 5060 Ti | llama 7B | Q4_0 | llama.cpp (CUDA) | 94.3 | 4034.65 | – | – | tg128 | 2025-04-24 | community | – | source |
| M2 Ultra (76 GPU) | llama 7B | Q4_0 | llama.cpp (Metal) | 94.27 | 1238.48 | – | – | – | 2023-11-22 | community | – | source |
| GeForce RTX 5060 Ti | llama 7B | Q4_0 | llama.cpp (CUDA) | 93.9 | 3912.3 | – | – | tg128 | 2026-03-06 | community | – | source |
| NVIDIA RTX 4500 | llama 7B | Q4_0 | llama.cpp (CUDA) | 93.88 | 6490.28 | – | – | tg128 | 2026-01-22 | community | – | source |
| Radeon 8060S | llama 7B | Q4_0 | llama.cpp (ROCm) | 93.84 | 4443.54 | – | – | tg128 | 2025-11-26 | community | – | source |
| RTX 5060 Ti | llama 7B | Q4_0 | llama.cpp (Vulkan) | 93.76 | 3713.27 | – | – | tg128 | 2025-11-04 | community | contradiction | source |
| RX 6800 | llama 7B | Q4_0 | llama.cpp (ROCm) | 93.6 | 1738.62 | – | – | tg128 | 2026-10-01 | community | – | source |
| Radeon RX 9060 XT | llama 7B | Q4_0 | llama.cpp (Vulkan) | 93.51 | 3460.92 | – | – | tg128 | 2026-02-03 | community | – | source |
| GeForce RTX 5060 Ti | llama 7B | Q4_0 | llama.cpp (CUDA) | 93.46 | 4195.53 | – | – | tg128 | 2025-08-02 | community | – | source |
| GeForce RTX 3080 | llama 7B | Q4_0 | llama.cpp (CUDA) | 93.42 | 3275.48 | – | – | tg128 | 2026-08-08 | community | – | source |
| RTX 5060 Ti | llama 7B | Q4_0 | llama.cpp (CUDA) | 93.3 | 4163.39 | – | – | tg128 | 2025-07-30 | community | – | source |
| Radeon RX 9070 XT | llama 7B | Q4_0 | llama.cpp (ROCm) | 93.26 | 4055.36 | – | – | tg128 | 2025-10-05 | community | – | source |
| Radeon RX 6900 XT | llama 7B | Q4_0 | llama.cpp (ROCm) | 92.96 | 1869.09 | – | – | tg128 | 2025-05-12 | community | – | source |
| Radeon RX 6950 XT | llama 7B | Q4_0 | llama.cpp (ROCm) | 92.96 | 1869.09 | – | – | tg128 | 2025-05-12 | community | – | source |
| NVIDIA RTX 4500 | llama 7B | Q4_0 | llama.cpp (CUDA) | 92.94 | 5512.71 | – | – | tg128 | 2026-01-22 | community | – | source |
| GeForce RTX 5060 Ti | llama 7B | Q4_0 | llama.cpp (CUDA) | 92.75 | 3541.12 | – | – | tg128 | 2026-05-08 | community | – | source |
| RTX PRO 4000 Blackwell | llama 7B | Q4_0 | llama.cpp (CUDA) | 92.54 | 4078.52 | – | – | tg128 | 2026-05-13 | community | – | source |
| GeForce RTX 5060 Ti | llama 7B | Q4_0 | llama.cpp (CUDA) | 92.52 | 3733.17 | – | – | tg128 | 2026-05-08 | community | – | source |
| GeForce RTX 3080 | llama 7B | Q4_0 | llama.cpp (CUDA) | 92.45 | 3211.59 | – | – | tg128 | 2025-12-22 | community | – | source |
| Radeon RX 7900 XTX | llama 7B | Q4_0 | llama.cpp (ROCm) | 92.28 | 1520.24 | – | – | tg128 | 2026-03-01 | community | – | source |
| GeForce RTX 3060 Ti | llama 7B | Q4_0 | llama.cpp (CUDA) | 92.23 | 2675.22 | – | – | tg128 | 2026-09-19 | community | – | source |
| RTX 5060 Ti | llama 7B | Q4_0 | llama.cpp (Vulkan) | 92.19 | 3342.62 | – | – | tg128 | 2026-02-03 | community | – | source |
| M3 Ultra (80 GPU) | llama 7B | Q4_0 | llama.cpp (Metal) | 92.14 | 1471.24 | – | – | – | 2023-11-22 | community | – | source |
| Radeon RX 9070 XT | llama 7B | Q4_0 | llama.cpp (ROCm) | 91.86 | 589.16 | – | – | tg128 | 2025-08-07 | community | – | source |
| RTX 5060 Ti | llama 7B | Q4_0 | llama.cpp (Vulkan) | 91.85 | 3720.85 | – | – | tg128 | 2025-11-04 | community | contradiction | source |
| V620 | llama 7B | Q4_0 | llama.cpp (ROCm) | 91.70 | 1846.20 | – | – | tg128 | 2026-09-10 | community | – | source |
| Radeon RX 6800 XT | llama 7B | Q4_0 | llama.cpp (Vulkan) | 91.59 | 863.03 | – | – | tg128 | 2025-01-08 | community | – | source |
| RTX 5060 Ti | llama 7B | Q4_0 | llama.cpp (CUDA) | 91.56 | 4073.37 | – | – | tg128 | 2025-07-30 | community | – | source |
| Radeon RX 7900 XTX | llama 7B | Q4_0 | llama.cpp (Vulkan) | 91.48 | 1484.04 | – | – | tg128 | 2025-01-08 | community | contradiction | source |
| RTX 5060 Ti | llama 7B | Q4_0 | llama.cpp (Vulkan) | 91.41 | 3240.12 | – | – | tg128 | 2025-11-04 | community | contradiction | source |
| RTX 5060 Ti | llama 7B | Q4_0 | llama.cpp (CUDA) | 90.95 | 3728.09 | – | – | tg128 | 2025-07-30 | community | – | source |
| GeForce RTX 5060 Ti | llama 7B | Q4_0 | llama.cpp (CUDA) | 90.94 | 3737.25 | – | – | tg128 | 2025-08-02 | community | – | source |
| GeForce RTX 3080 | llama 7B | Q4_0 | llama.cpp (CUDA) | 90.78 | 2970.97 | 149.0 | – | tg128 | 2025-12-22 | community | – | source |
| GeForce RTX 5060 Ti | llama 7B | Q4_0 | llama.cpp (Vulkan) | 90.61 | 3359.51 | – | – | tg128 | 2026-05-08 | community | – | source |
| GeForce RTX 3080 | llama 7B | Q4_0 | llama.cpp (CUDA) | 90.6 | 2875.02 | – | – | tg128 | 2026-08-08 | community | – | source |
| RTX 3060 Ti | llama 7B | Q4_0 | llama.cpp (Vulkan) | 90.35 | 2644.95 | – | – | tg128 | 2026-04-30 | community | – | source |
| Radeon RX 9070 XT | llama 7B | Q4_0 | llama.cpp (ROCm) | 90.08 | 723.41 | – | – | tg128 | 2025-08-07 | community | – | source |
| Radeon RX 9070 XT | llama 7B | Q4_0 | llama.cpp (ROCm) | 89.9 | 4065.43 | – | – | tg128 | 2025-09-23 | community | – | source |
| RTX 5060 Ti | llama 7B | Q4_0 | llama.cpp (CUDA) | 89.87 | 3917.29 | – | – | tg128 | 2025-07-30 | community | – | source |
| RTX 5060 Ti | llama 7B | Q4_0 | llama.cpp (CUDA) | 89.75 | 3526.38 | – | – | tg128 | 2025-07-30 | community | – | source |
| RTX PRO 4000 Blackwell | llama 7B | Q4_0 | llama.cpp (CUDA) | 89.73 | 3628.68 | – | – | tg128 | 2026-05-13 | community | – | source |
| RTX 5060 Ti | llama 7B | Q4_0 | llama.cpp (CUDA) | 89.57 | 3648.71 | – | – | tg128 | 2025-07-30 | community | – | source |
| Radeon RX 9070 XT | llama 7B | Q4_0 | llama.cpp (ROCm) | 89.51 | 4543.73 | – | – | tg128 | 2025-11-08 | community | – | source |
| Radeon RX 9070 XT | llama 7B | Q4_0 | llama.cpp (ROCm) | 89.17 | 4027.14 | – | – | tg128 | 2025-09-23 | community | – | source |
| Radeon RX 9070 XT | llama 7B | Q4_0 | llama.cpp (ROCm) | 88.96 | 4029.45 | – | – | tg128 | 2025-10-05 | community | – | source |
| M2 Ultra (60 GPU) | llama 7B | Q4_0 | llama.cpp (Metal) | 88.64 | 1013.81 | – | – | – | 2023-11-22 | community | – | source |
| Radeon RX 6900 XT | llama 7B | Q4_0 | llama.cpp (ROCm) | 88.49 | 1889.84 | – | – | tg128 | 2025-09-09 | community | – | source |
| M3 Ultra (60 GPU) | llama 7B | Q4_0 | llama.cpp (Metal) | 88.4 | 1073.09 | – | – | – | 2023-11-22 | community | – | source |
| Radeon RX 6800 XT | llama 7B | Q4_0 | llama.cpp (ROCm) | 88.31 | 1706.8 | – | – | tg128 | 2025-09-01 | community | – | source |
| Arc B580 | llama 7B | Q4_0 | llama.cpp (SYCL) | 88.13 | 893.17 | – | – | tg128 | 2026-06-13 | community | – | source |
| Radeon RX 7900 XTX | llama 7B | Q4_0 | llama.cpp (Vulkan) | 88.12 | 1545.39 | – | – | tg128 | 2025-01-08 | community | contradiction | source |
| Radeon RX 9070 XT | llama 7B | Q4_0 | llama.cpp (ROCm) | 87.57 | 4131.22 | – | – | tg128 | 2025-09-03 | community | – | source |
| Radeon RX 6800 XT | llama 7B | Q4_0 | llama.cpp (Vulkan) | 87.56 | 519.14 | – | – | tg128 | 2025-05-07 | community | – | source |
| Radeon RX 9070 XT | llama 7B | Q4_0 | llama.cpp (ROCm) | 87.43 | 3538.39 | – | – | tg128 | 2025-09-03 | community | – | source |
| RTX 5060 Ti | llama 7B | Q4_0 | llama.cpp (Vulkan) | 87.05 | 3248.01 | – | – | tg128 | 2025-11-04 | community | contradiction | source |
| Radeon RX 6800 | llama 7B | Q4_0 | llama.cpp (ROCm) | 86.91 | 1675.31 | – | – | tg128 | 2026-06-04 | community | – | source |
| RTX 3060 Ti | llama 7B | Q4_0 | llama.cpp (Vulkan) | 86.74 | 2361.43 | – | – | tg128 | 2026-04-30 | community | – | source |
| Arc B580 | llama 7B | Q4_0 | llama.cpp (SYCL) | 86.54 | 1806.18 | – | – | tg128 | 2026-06-04 | community | – | source |
| Radeon RX 6800 XT | llama 7B | Q4_0 | llama.cpp (Vulkan) | 86.47 | 510.8 | – | – | tg128 | 2025-05-07 | community | – | source |
| Radeon RX 6800 XT | llama 7B | Q4_0 | llama.cpp (Vulkan) | 86.17 | 522.39 | – | – | tg128 | 2025-05-07 | community | – | source |
| Radeon RX 9070 XT | llama 7B | Q4_0 | llama.cpp (ROCm) | 86.16 | 3750.46 | – | – | tg128 | 2025-11-08 | community | – | source |
| Radeon RX 9070 XT | llama 7B | Q4_0 | llama.cpp (ROCm) | 86.1 | 2909.86 | – | – | tg128 | 2025-05-26 | community | – | source |
| RX 6800 | llama 7B | Q4_0 | llama.cpp (ROCm) | 86.03 | 1510.13 | – | – | tg128 | 2026-10-01 | community | – | source |
| Radeon RX 6900 XT | llama 7B | Q4_0 | llama.cpp (ROCm) | 85.6 | 1843.15 | – | – | tg128 | 2025-05-12 | community | – | source |
| Radeon RX 6900 XT | llama 7B | Q4_0 | llama.cpp (ROCm) | 85.6 | 1843.15 | – | – | tg128 | 2025-05-12 | community | – | source |
| Radeon RX 6950 XT | llama 7B | Q4_0 | llama.cpp (ROCm) | 85.49 | 1956.82 | – | – | tg128 | 2025-05-12 | community | – | source |
| Radeon RX 6900 XT | llama 7B | Q4_0 | llama.cpp (ROCm) | 85.48 | 1963.96 | – | – | tg128 | 2025-01-18 | community | – | source |
| Radeon RX 9070 XT | llama 7B | Q4_0 | llama.cpp (ROCm) | 85.43 | 1341.03 | – | – | tg128 | 2025-08-24 | community | – | source |
| Radeon RX 6900 XT | llama 7B | Q4_0 | llama.cpp (ROCm) | 85.42 | 1962.01 | – | – | tg128 | 2025-05-12 | community | – | source |
| A4000 | llama 7B | Q4_0 | llama.cpp (CUDA) | 85.17 | 2807.83 | – | – | tg128 | 2025-08-17 | community | – | source |
| Radeon RX 9070 XT | llama 7B | Q4_0 | llama.cpp (ROCm) | 85.12 | 2738.43 | – | – | tg128 | 2025-05-26 | community | – | source |
| Radeon RX 6900 XT | llama 7B | Q4_0 | llama.cpp (ROCm) | 85.04 | 1948.31 | – | – | tg128 | 2025-09-09 | community | – | source |
| Radeon RX 6800 XT | llama 7B | Q4_0 | llama.cpp (ROCm) | 84.88 | 1199.27 | – | – | tg128 | 2025-09-01 | community | – | source |
| RTX 3070 | llama 7B | Q4_0 | llama.cpp (CUDA) | 84.62 | 2993.73 | – | – | tg128 | 2025-05-29 | community | – | source |
| Arc Pro B60 | llama 7B | Q4_0 | llama.cpp (SYCL) | 84.44 | 928.91 | – | – | tg128 | 2026-08-27 | community | – | source |
| Radeon RX 9070 XT | llama 7B | Q4_0 | llama.cpp (ROCm) | 84.34 | 2728.17 | – | – | tg128 | 2025-05-26 | community | – | source |
| Arc Pro B60 | llama 7B | Q4_0 | llama.cpp (SYCL) | 84.12 | 664.82 | – | – | tg128 | 2026-08-27 | community | – | source |
| Radeon RX 6800 | llama 7B | Q4_0 | llama.cpp (ROCm) | 83.92 | 1447.07 | – | – | tg128 | 2025-08-11 | community | – | source |
| Radeon RX 6700 XT | llama 7B | Q4_0 | llama.cpp (Vulkan) | 83.88 | 1051.2 | – | – | tg128 | 2025-09-16 | community | – | source |
| A4000 | llama 7B | Q4_0 | llama.cpp (CUDA) | 83.77 | 2684.06 | – | – | tg128 | 2025-08-17 | community | – | source |
| M1 Ultra (64 GPU) | llama 7B | Q4_0 | llama.cpp (Metal) | 83.73 | 1030.04 | – | – | – | 2023-11-22 | community | – | source |
| Arc B580 | llama 7B | Q4_0 | llama.cpp (SYCL) | 83.67 | 1954.15 | – | – | tg128 | 2026-06-13 | community | – | source |
| Radeon RX 9070 XT | llama 7B | Q4_0 | llama.cpp (ROCm) | 83.63 | 2908.66 | – | – | tg128 | 2025-08-24 | community | – | source |
| RTX 3090 | llama 7B | Q4_0 | llama.cpp (ROCm) | 83.55 | 997.21 | – | – | tg128 | 2025-09-01 | community | – | source |
| Radeon RX 6900 XT | llama 7B | Q4_0 | llama.cpp (Vulkan) | 83.54 | 713.77 | – | – | tg128 | 2025-02-08 | community | – | source |
| Arc B580 | llama 7B | Q4_0 | llama.cpp (SYCL) | 83.39 | 1542.23 | – | – | tg128 | 2026-06-04 | community | – | source |
| RTX 5060 Ti | llama 7B | Q4_0 | llama.cpp (Vulkan) | 83.26 | 3492.22 | – | – | tg128 | 2025-04-22 | community | – | source |
| M4 Max (40 GPU) | llama 7B | Q4_0 | llama.cpp (Metal) | 83.06 | 885.68 | – | – | – | 2023-11-22 | community | – | source |
| Radeon RX 6900 XT | llama 7B | Q4_0 | llama.cpp (ROCm) | 83.02 | 1824.47 | – | – | tg128 | 2025-08-15 | community | – | source |
| A770 | llama 7B | Q4_0 | llama.cpp (SYCL) | 82.92 | 2416.16 | – | – | tg128 | 2025-09-22 | community | – | source |
| Arc B580 | llama 7B | Q4_0 | llama.cpp (SYCL) | 82.85 | 1631.3 | – | – | tg128 | 2026-06-04 | community | – | source |
| Radeon RX 6950 XT | llama 7B | Q4_0 | llama.cpp (ROCm) | 82.73 | 1666.05 | – | – | tg128 | 2025-05-12 | community | – | source |
| Arc B580 | llama 7B | Q4_0 | llama.cpp (SYCL) | 82.49 | 2015.51 | – | – | tg128 | 2026-06-04 | community | – | source |
| Radeon RX 6750 XT | llama 7B | Q4_0 | llama.cpp (Vulkan) | 82.29 | 997.05 | – | – | tg128 | 2025-06-10 | community | – | source |
| Radeon RX 6900 XT | llama 7B | Q4_0 | llama.cpp (Vulkan) | 82.26 | 643.37 | – | – | tg128 | 2025-08-11 | community | – | source |
| Arc Pro B70 | llama 7B | Q4_0 | llama.cpp (Vulkan) | 82.19 | 2461.92 | – | – | tg128 | 2026-07-17 | community | – | source |
| Radeon RX 6750 XT | llama 7B | Q4_0 | llama.cpp (Vulkan) | 81.98 | 1040.58 | – | – | tg128 | 2025-06-10 | community | – | source |
| Radeon RX 6700 XT | llama 7B | Q4_0 | llama.cpp (Vulkan) | 81.86 | 1010.9 | – | – | tg128 | 2025-09-16 | community | – | source |
| A770 | llama 7B | Q4_0 | llama.cpp (SYCL) | 81.6 | 2417.17 | – | – | tg128 | 2025-09-22 | community | – | source |
| RTX 5060 Ti | llama 7B | Q4_0 | llama.cpp (Vulkan) | 81.48 | 3211.73 | – | – | tg128 | 2025-04-22 | community | – | source |
| NVIDIA RTX 3500 | llama 7B | Q4_0 | llama.cpp (CUDA) | 81.17 | 3353.26 | – | – | tg128 | 2025-09-20 | community | – | source |
| Radeon RX 9070 XT | llama 7B | Q4_0 | llama.cpp (ROCm) | 80.81 | 1841.07 | – | – | tg128 | 2025-05-26 | community | – | source |
| RTX 3060 | llama 7B | Q4_0 | llama.cpp (Vulkan) | 80.59 | 2012.88 | – | – | tg128 | 2025-11-23 | community | – | source |
| Radeon RX 6900 XT | llama 7B | Q4_0 | llama.cpp (ROCm) | 80.45 | 1250.68 | – | – | tg128 | 2025-08-15 | community | – | source |
| Arc Pro B70 | llama 7B | Q4_0 | llama.cpp (Vulkan) | 80.39 | 2531.36 | – | – | tg128 | 2026-07-17 | community | – | source |
| A4000 | llama 7B | Q4_0 | llama.cpp (CUDA) | 80.39 | 2835.78 | – | – | tg128 | 2026-01-08 | community | – | source |
| NVIDIA RTX 3500 | llama 7B | Q4_0 | llama.cpp (CUDA) | 80.18 | 3757.32 | – | – | tg128 | 2025-09-20 | community | – | source |
| Arc Pro B60 | llama 7B | Q4_0 | llama.cpp (SYCL) | 80.03 | 1869.16 | – | – | tg128 | 2026-08-27 | community | – | source |
| Arc Pro B60 | llama 7B | Q4_0 | llama.cpp (SYCL) | 80.03 | 570.18 | – | – | tg128 | 2026-08-27 | community | – | source |
| Radeon RX 6800 | llama 7B | Q4_0 | llama.cpp (ROCm) | 79.78 | 1382.74 | – | – | tg128 | 2026-06-04 | community | – | source |
| Radeon RX 6800 XT | llama 7B | Q4_0 | llama.cpp (Vulkan) | 79.6 | 444.07 | – | – | tg128 | 2025-05-07 | community | contradiction | source |
| Radeon RX 6800 XT | llama 7B | Q4_0 | llama.cpp (Vulkan) | 79.35 | 441.9 | – | – | tg128 | 2025-05-07 | community | contradiction | source |
| M1 Ultra | llama 7B | Q8_0 | llama.cpp (Metal) | 79.13 | 1167.67 | – | – | tg128 | 2026-08-28 | community | – | source |
| RTX 3070 | llama 7B | Q4_0 | llama.cpp (Vulkan) | 78.71 | 2113.02 | – | – | tg128 | 2025-05-29 | community | – | source |
| Radeon RX 6800 XT | llama 7B | Q4_0 | llama.cpp (Vulkan) | 78.3 | 570.06 | – | – | tg128 | 2025-05-07 | community | contradiction | source |
| Radeon RX 6750 XT | llama 7B | Q4_0 | llama.cpp (Vulkan) | 78.19 | 920.41 | – | – | tg128 | 2026-07-20 | community | – | source |
| GeForce RTX 3060 | llama 7B | Q4_0 | llama.cpp (CUDA) | 78.19 | 2452.67 | – | – | tg128 | 2025-11-23 | community | – | source |
| Arc B580 | llama 7B | Q4_0 | llama.cpp (SYCL) | 78.15 | 1737.0 | – | – | tg128 | 2026-06-04 | community | – | source |
| A4000 | llama 7B | Q4_0 | llama.cpp (CUDA) | 78.12 | 2523.51 | – | – | tg128 | 2026-01-08 | community | – | source |
| RTX 5060 Ti | llama 7B | Q4_0 | llama.cpp (Vulkan) | 78.1 | 3504.54 | – | – | tg128 | 2025-07-30 | community | – | source |
| Arc B580 | llama 7B | Q4_0 | llama.cpp (SYCL) | 77.92 | 1851.35 | – | – | tg128 | 2026-06-04 | community | – | source |
| Radeon RX 6900 XT | llama 7B | Q4_0 | llama.cpp (ROCm) | 77.84 | 1615.12 | – | – | tg128 | 2025-05-12 | community | – | source |
| Quadro RTX 5000 | llama 7B | Q4_0 | llama.cpp (CUDA) | 77.74 | 1970.1 | – | – | tg128 | 2026-05-25 | community | – | source |
| Arc Pro B60 | llama 7B | Q4_0 | llama.cpp (SYCL) | 77.65 | 652.15 | 152.0 | – | tg128 | 2026-01-17 | community | – | source |
| Radeon RX 7900 XT | llama 7B | Q4_0 | llama.cpp (Vulkan) | 77.0 | 172.66 | – | – | tg128 | 2025-09-24 | community | – | source |
| GeForce RTX 3060 | llama 7B | Q4_0 | llama.cpp (CUDA) | 76.92 | 2407.67 | – | – | tg128 | 2025-08-21 | community | – | source |
| Arc Pro B60 | llama 7B | Q4_0 | llama.cpp (Vulkan) | 76.83 | 1841.31 | – | – | tg128 | 2026-06-04 | community | – | source |
| A4000 | llama 7B | Q4_0 | llama.cpp (Vulkan) | 76.07 | 2482.74 | – | – | tg128 | 2026-01-08 | community | – | source |
| RTX 3060 | llama 7B | Q4_0 | llama.cpp (Vulkan) | 75.94 | 1815.7 | – | – | tg128 | 2025-11-23 | community | – | source |
| Radeon RX 6950 XT | llama 7B | Q4_0 | llama.cpp (ROCm) | 75.8 | 1730.22 | – | – | tg128 | 2025-05-12 | community | – | source |
| Arc A770 | llama 7B | Q4_0 | llama.cpp (Vulkan) | 75.75 | 1479.07 | – | – | tg128 | 2025-07-30 | community | – | source |
| GeForce RTX 3060 | llama 7B | Q4_0 | llama.cpp (CUDA) | 75.58 | 2407.93 | – | – | tg128 | 2026-08-21 | community | – | source |
| GeForce RTX 3060 | llama 7B | Q4_0 | llama.cpp (CUDA) | 75.57 | 2137.5 | – | – | tg128 | 2025-08-21 | community | – | source |
| Radeon RX 6900 XT | llama 7B | Q4_0 | llama.cpp (ROCm) | 75.24 | 1736.24 | – | – | tg128 | 2025-05-12 | community | – | source |
| GeForce RTX 3060 | llama 7B | Q4_0 | llama.cpp (CUDA) | 75.17 | 2118.17 | – | – | tg128 | 2025-11-23 | community | – | source |
| M4 Max | llama 7B | Q4_0 | llama.cpp (Vulkan) | 75.02 | 724.77 | – | – | tg128 | 2025-07-15 | community | – | source |
| GeForce RTX 3060 | llama 7B | Q4_0 | llama.cpp (CUDA) | 74.98 | 2401.11 | – | – | tg128 | 2025-08-23 | community | – | source |
| M1 Ultra (48 GPU) | llama 7B | Q4_0 | llama.cpp (Metal) | 74.93 | 772.24 | – | – | – | 2023-11-22 | community | – | source |
| RTX 5060 Ti | llama 7B | Q4_0 | llama.cpp (Vulkan) | 74.8 | 3386.08 | – | – | tg128 | 2025-07-30 | community | – | source |
| Quadro RTX 5000 | llama 7B | Q4_0 | llama.cpp (CUDA) | 74.76 | 1840.52 | – | – | tg128 | 2026-05-25 | community | – | source |
| A5500 | llama 7B | Q4_0 | llama.cpp (CUDA) | 74.68 | 2180.23 | – | – | tg128 | 2026-05-25 | community | – | source |
| A5500 | llama 7B | Q4_0 | llama.cpp (CUDA) | 74.66 | 2338.14 | – | – | tg128 | 2026-05-25 | community | – | source |
| Radeon RX 7900 XTX | llama 7B | Q4_0 | llama.cpp (ROCm) | 74.66 | 1803.65 | – | – | tg128 | 2025-08-02 | community | – | source |
| RTX 5060 Ti | llama 7B | Q4_0 | llama.cpp (Vulkan) | 74.52 | 3060.78 | – | – | tg128 | 2025-07-30 | community | – | source |
| RTX 5060 Ti | llama 7B | Q4_0 | llama.cpp (Vulkan) | 74.47 | 3226.8 | – | – | tg128 | 2025-07-30 | community | – | source |
| Radeon RX 5700 XT | llama 7B | Q4_0 | llama.cpp (Vulkan) | 74.43 | 538.31 | – | – | tg128 | 2026-01-29 | community | – | source |
| Arc A770 | llama 7B | Q4_0 | llama.cpp (Vulkan) | 74.24 | 1543.17 | – | – | tg128 | 2025-07-30 | community | – | source |
| Radeon RX 9060 XT | llama 7B | Q4_0 | llama.cpp (Vulkan) | 74.01 | 3228.15 | – | – | tg128 | 2026-10-02 | community | – | source |
| Radeon RX 9060 XT | llama 7B | Q4_0 | llama.cpp (Vulkan) | 73.86 | 2311.69 | – | – | tg128 | 2026-02-28 | community | – | source |
| Arc B580 | llama 7B | Q4_0 | llama.cpp (SYCL) | 73.76 | 2063.52 | – | – | tg128 | 2026-05-27 | community | – | source |
| GeForce RTX 3060 | llama 7B | Q4_0 | llama.cpp (CUDA) | 73.75 | 2343.25 | – | – | tg128 | 2025-08-18 | community | – | source |
| A4000 | llama 7B | Q4_0 | llama.cpp (Vulkan) | 73.74 | 2248.14 | – | – | tg128 | 2026-01-08 | community | – | source |
| GeForce RTX 3060 | llama 7B | Q4_0 | llama.cpp (CUDA) | 73.74 | 2346.54 | – | – | tg128 | 2026-02-19 | community | – | source |
| RTX 3060 | llama 7B | Q4_0 | llama.cpp (Vulkan) | 73.71 | 1685.14 | – | – | tg128 | 2025-12-03 | community | – | source |
| M5 Ultra | llama 7B 6.74B | F16 | llama.cpp (Metal) | 73.6 | 5207.14 | – | – | tg128 | 2026-09-30 | community | – | source |
| RTX 3090 | llama 7B | Q4_0 | llama.cpp (ROCm) | 73.58 | 426.56 | – | – | tg128 | 2025-09-01 | community | – | source |
| Arc Pro B60 | llama 7B | Q4_0 | llama.cpp (Vulkan) | 73.01 | 1929.93 | – | – | tg128 | 2026-06-04 | community | – | source |
| A770 | llama 7B | Q4_0 | llama.cpp (SYCL) | 72.54 | 2564.25 | – | – | tg128 | 2025-09-21 | community | – | source |
| M5 Max | llama 7B | Q8_0 | llama.cpp (Metal) | 72.52 | 3219.94 | – | – | tg128 | 2026-06-22 | community | – | source |
| M5 Max (40 GPU) | llama 7B | Q8_0 | llama.cpp (Metal) | 72.42 | 3143.81 | – | – | – | 2023-11-22 | community | – | source |
| M5 Max | llama 7B | Q8_0 | llama.cpp (Metal) | 72.42 | 3143.81 | – | – | tg128 | 2026-08-25 | community | – | source |
| Radeon RX 6900 XT | llama 7B | Q4_0 | llama.cpp (Vulkan) | 72.39 | 111.82 | – | – | tg128 | 2025-10-26 | community | – | source |
| Radeon RX 6900 XT | llama 7B | Q4_0 | llama.cpp (Vulkan) | 72.31 | 118.09 | – | – | tg128 | 2025-10-26 | community | – | source |
| M4 | llama 7B | Q4_0 | llama.cpp (Metal) | 72.26 | 762.81 | – | – | tg128 | 2025-05-20 | community | – | source |
| Radeon RX 9060 XT | llama 7B | Q4_0 | llama.cpp (Vulkan) | 72.24 | 1840.64 | – | – | tg128 | 2025-09-14 | community | – | source |
| GeForce RTX 3060 | llama 7B | Q4_0 | llama.cpp (CUDA) | 72.09 | 2111.39 | – | – | tg128 | 2026-08-21 | community | – | source |
| Radeon RX 9060 XT | llama 7B | Q4_0 | llama.cpp (Vulkan) | 72.0 | 411.98 | – | – | tg128 | 2026-02-28 | community | – | source |
| Arc B580 | llama 7B | Q4_0 | llama.cpp (Vulkan) | 72.0 | 419.49 | – | – | tg128 | 2025-09-20 | community | – | source |
| Radeon RX 6900 XT | llama 7B | Q4_0 | llama.cpp (Vulkan) | 71.96 | 116.28 | – | – | tg128 | 2025-10-26 | community | – | source |
| Radeon RX 9060 XT | llama 7B | Q4_0 | llama.cpp (Vulkan) | 71.95 | 2056.43 | – | – | tg128 | 2025-09-14 | community | – | source |
| Arc A770 | llama 7B | Q4_0 | llama.cpp (ROCm) | 71.93 | 1357.35 | – | – | tg128 | 2025-07-30 | community | – | source |
| M4 Max | llama 7B | Q4_0 | llama.cpp (Vulkan) | 71.79 | 557.46 | – | – | tg128 | 2025-07-15 | community | – | source |
| Radeon RX 5700 XT | llama 7B | Q4_0 | llama.cpp (Vulkan) | 71.66 | 474.89 | – | – | tg128 | 2025-09-23 | community | – | source |
| GeForce RTX 3060 | llama 7B | Q4_0 | llama.cpp (CUDA) | 71.66 | 1921.04 | – | – | tg128 | 2025-08-18 | community | – | source |
| Radeon RX 5700 XT | llama 7B | Q4_0 | llama.cpp (Vulkan) | 71.58 | 492.13 | – | – | tg128 | 2025-09-23 | community | – | source |
| Radeon RX 7900 XT | llama 7B | Q4_0 | llama.cpp (Vulkan) | 71.52 | 92.18 | – | – | tg128 | 2025-09-24 | community | – | source |
| Radeon RX 9060 XT | llama 7B | Q4_0 | llama.cpp (Vulkan) | 71.45 | 2093.56 | – | – | tg128 | 2026-02-28 | community | – | source |
| Arc Pro B60 | llama 7B | Q4_0 | llama.cpp (Vulkan) | 71.42 | 1572.29 | – | – | tg128 | 2026-06-04 | community | – | source |
| 2x V620 | llama 7B | Q4_0 | llama.cpp (ROCm) | 71.38 | 2146.36 | – | – | tg128 | 2026-09-10 | community | – | source |
| Radeon RX 6800 XT | llama 7B | Q4_0 | llama.cpp (Vulkan) | 71.37 | 612.47 | – | – | tg128 | 2025-04-12 | community | – | source |
| A770 | llama 7B | Q4_0 | llama.cpp (SYCL) | 71.32 | 2545.16 | – | – | tg128 | 2025-09-21 | community | – | source |
| GeForce RTX 3060 | llama 7B | Q4_0 | llama.cpp (CUDA) | 71.17 | 2239.81 | – | – | tg128 | 2025-08-18 | community | – | source |
| GeForce RTX 3060 | llama 7B | Q4_0 | llama.cpp (CUDA) | 71.14 | 2020.41 | – | – | tg128 | 2025-08-23 | community | – | source |
| Radeon RX 9060 XT | llama 7B | Q4_0 | llama.cpp (Vulkan) | 71.07 | 2844.8 | – | – | tg128 | 2026-10-02 | community | – | source |
| Radeon RX 7900 XTX | llama 7B | Q4_0 | llama.cpp (ROCm) | 70.83 | 1256.86 | – | – | tg128 | 2025-08-02 | community | – | source |
| RTX 3060 | llama 7B | Q4_0 | llama.cpp (CUDA) | 70.78 | 2229.42 | – | – | tg128 | 2025-06-17 | community | – | source |
| Radeon RX 5700 XT | llama 7B | Q4_0 | llama.cpp (Vulkan) | 70.73 | 529.69 | – | – | tg128 | 2026-01-29 | community | – | source |
| Arc Pro B60 | llama 7B | Q4_0 | llama.cpp (Vulkan) | 70.54 | 274.76 | – | – | tg128 | 2026-01-17 | community | – | source |
| A770 | llama 7B | Q4_0 | llama.cpp (SYCL) | 70.54 | 2498.81 | – | – | tg128 | 2026-01-28 | community | – | source |
| Radeon RX 9060 XT | llama 7B | Q4_0 | llama.cpp (Vulkan) | 70.54 | 2141.67 | – | – | tg128 | 2025-06-13 | community | – | source |
| Radeon RX 9060 XT | llama 7B | Q4_0 | llama.cpp (Vulkan) | 70.52 | 1915.41 | – | – | tg128 | 2025-06-13 | community | – | source |
| GeForce RTX 3060 | llama 7B | Q4_0 | llama.cpp (CUDA) | 70.52 | 2070.83 | – | – | tg128 | 2026-02-19 | community | – | source |
| A770 | llama 7B | Q4_0 | llama.cpp (SYCL) | 70.18 | 2478.2 | – | – | tg128 | 2025-04-01 | community | – | source |
| Radeon RX 9060 XT | llama 7B | Q4_0 | llama.cpp (ROCm) | 70.18 | 3014.51 | – | – | tg128 | 2026-10-02 | community | – | source |
| Radeon RX 6800 XT | llama 7B | Q4_0 | llama.cpp (Vulkan) | 70.17 | 312.02 | – | – | tg128 | 2025-01-08 | community | – | source |
| Arc B580 | llama 7B | Q4_0 | llama.cpp (Vulkan) | 70.14 | 620.94 | – | – | tg128 | 2025-09-20 | community | – | source |
| Radeon RX 5700 XT | llama 7B | Q4_0 | llama.cpp (Vulkan) | 70.13 | 439.42 | – | – | tg128 | 2025-01-12 | community | – | source |
| M4 Max (32 GPU) | llama 7B | Q4_0 | llama.cpp (Metal) | 69.95 | 713.93 | – | – | – | 2023-11-22 | community | – | source |
| A770 | llama 7B | Q4_0 | llama.cpp (SYCL) | 69.73 | 2512.42 | – | – | tg128 | 2026-01-28 | community | outlier | source |
| Arc Pro B60 | llama 7B | Q4_0 | llama.cpp (Vulkan) | 69.46 | 1626.71 | – | – | tg128 | 2026-06-04 | community | – | source |
| Arc A770 | llama 7B | Q4_0 | llama.cpp (ROCm) | 69.27 | 987.41 | – | – | tg128 | 2025-07-30 | community | – | source |
| Radeon RX 5700 XT | llama 7B | Q4_0 | llama.cpp (Vulkan) | 69.14 | 605.4 | – | – | tg128 | 2025-01-15 | community | – | source |
| GeForce RTX 3060 | llama 7B | Q4_0 | llama.cpp (CUDA) | 69.01 | 2028.81 | – | – | tg128 | 2025-08-18 | community | – | source |
| Radeon 8060S | llama 7B | Q4_0 | llama.cpp (ROCm) | 68.67 | 4727.48 | – | – | tg128 | 2025-11-26 | community | contradiction | source |
| Arc Pro B60 | llama 7B | Q4_0 | llama.cpp (Vulkan) | 68.55 | 522.36 | 126 | – | tg128 | 2026-01-17 | community | – | source |
| RTX 3060 | llama 7B | Q4_0 | llama.cpp (CUDA) | 68.38 | 2020.4 | – | – | tg128 | 2025-06-17 | community | – | source |
| RTX 3060 | llama 7B | Q4_0 | llama.cpp (Vulkan) | 68.34 | 1122.04 | – | – | tg128 | 2025-12-03 | community | – | source |
| Arc Pro B60 | llama 7B | Q4_0 | llama.cpp (Vulkan) | 68.07 | 1621.0 | – | – | tg128 | 2026-06-04 | community | contradiction | source |
| Radeon RX 7600 XT | llama 7B | Q4_0 | llama.cpp (Vulkan) | 67.98 | 2251.85 | – | – | tg128 | 2026-02-15 | community | – | source |
| Quadro RTX 4000 | llama 7B | Q4_0 | llama.cpp (CUDA) | 67.62 | 1662.8 | – | – | tg128 | 2025-11-21 | community | – | source |
| M5 Pro | llama 7B | Q4_0 | llama.cpp (Metal) | 67.59 | 1340.44 | – | – | tg128 | 2026-09-24 | community | – | source |
| Radeon RX 9060 XT | llama 7B | Q4_0 | llama.cpp (ROCm) | 67.58 | 1419.67 | – | – | tg128 | 2025-09-14 | community | – | source |
| Radeon RX 5700 XT | llama 7B | Q4_0 | llama.cpp (ROCm) | 67.55 | 354.17 | – | – | tg128 | 2025-01-12 | community | – | source |
| Radeon RX 5700 XT | llama 7B | Q4_0 | llama.cpp (Vulkan) | 67.53 | 609.57 | – | – | tg128 | 2025-01-15 | community | – | source |
| Radeon RX 9060 XT | llama 7B | Q4_0 | llama.cpp (ROCm) | 67.27 | 2641.07 | – | – | tg128 | 2026-10-02 | community | – | source |
| RTX 3060 | llama 7B | Q4_0 | llama.cpp (Vulkan) | 67.11 | 1826.49 | – | – | tg128 | 2025-06-17 | community | – | source |
| Radeon RX 6900 XT | llama 7B | Q4_0 | llama.cpp (Vulkan) | 66.81 | 659.07 | – | – | tg128 | 2025-05-07 | community | – | source |
| Radeon RX 9060 XT | llama 7B | Q4_0 | llama.cpp (Vulkan) | 66.67 | 446.42 | – | – | tg128 | 2026-02-28 | community | – | source |
| M2 Ultra (76 GPU) | llama 7B | Q8_0 | llama.cpp (Metal) | 66.64 | 1248.59 | – | – | – | 2023-11-22 | community | – | source |
| GeForce RTX 5060 Ti | llama 7B | Q4_0 | llama.cpp (Vulkan) | 66.59 | 572.26 | – | – | tg128 | 2026-05-08 | community | contradiction | source |
| M2 | llama 7B | Q4_0 | llama.cpp (Vulkan) | 66.48 | 685.51 | – | – | tg128 | 2025-11-26 | community | – | source |
| Radeon RX 9060 XT | llama 7B | Q4_0 | llama.cpp (ROCm) | 66.48 | 2534.4 | – | – | tg128 | 2025-12-15 | community | – | source |
| M5 Pro (20 GPU) | llama 7B | Q4_0 | llama.cpp (Metal) | 66.33 | 1620.64 | – | – | – | 2023-11-22 | community | – | source |
| M5 Pro | llama 7B | Q4_0 | llama.cpp (Metal) | 66.33 | 1620.64 | – | – | tg128 | 2026-08-31 | community | – | source |
| M2 | llama 7B | Q4_0 | llama.cpp (Vulkan) | 66.32 | 683.61 | – | – | tg128 | 2025-11-28 | community | – | source |
| M3 Max (40 GPU) | llama 7B | Q4_0 | llama.cpp (Metal) | 66.31 | 759.7 | – | – | – | 2023-11-22 | community | – | source |
| Radeon 8060S | llama 7B | Q4_0 | llama.cpp (ROCm) | 66.27 | 3585.2 | – | – | tg128 | 2025-11-26 | community | contradiction | source |
| RTX 3060 | llama 7B | Q4_0 | llama.cpp (Vulkan) | 66.12 | 1715.17 | – | – | tg128 | 2025-08-23 | community | – | source |
| M2 Max (38 GPU) | llama 7B | Q4_0 | llama.cpp (Metal) | 65.95 | 671.31 | – | – | – | 2023-11-22 | community | – | source |
| Arc Pro B60 | llama 7B | Q4_0 | llama.cpp (Vulkan) | 65.94 | 1667.43 | – | – | tg128 | 2026-06-04 | community | – | source |
| A100 | llama 7B | Q4_0 | llama.cpp (CUDA) | 65.91 | 653.01 | – | – | tg128 | 2026-05-13 | community | contradiction | source |
| Arc B580 | llama 7B | Q4_0 | llama.cpp (SYCL) | 65.67 | 1721.91 | – | – | tg128 | 2026-05-27 | community | contradiction | source |
| RTX 3060 | llama 7B | Q4_0 | llama.cpp (Vulkan) | 65.63 | 1730.33 | – | – | tg128 | 2025-08-23 | community | – | source |
| Quadro RTX 4000 | llama 7B | Q4_0 | llama.cpp (CUDA) | 65.62 | 1536.89 | – | – | tg128 | 2025-11-21 | community | – | source |
| A100 | llama 7B | Q4_0 | llama.cpp (CUDA) | 65.48 | 626.32 | – | – | tg128 | 2026-05-13 | community | contradiction | source |
| Radeon RX 9060 XT | llama 7B | Q4_0 | llama.cpp (ROCm) | 65.42 | 1479.27 | – | – | tg128 | 2025-09-14 | community | – | source |
| Radeon RX 7600 XT | llama 7B | Q4_0 | llama.cpp (Vulkan) | 65.36 | 2037.24 | – | – | tg128 | 2026-02-15 | community | – | source |
| 2x V620 | llama 7B | Q4_0 | llama.cpp (ROCm) | 64.85 | 1801.75 | – | – | tg128 | 2026-09-10 | community | – | source |
| RTX 3060 | llama 7B | Q4_0 | llama.cpp (Vulkan) | 64.76 | 1681.81 | – | – | tg128 | 2025-06-17 | community | – | source |
| RTX 2070 Super | llama 7B | Q4_0 | llama.cpp (Vulkan) | 64.64 | 1199.13 | – | – | tg128 | 2025-02-07 | community | – | source |
| Radeon RX 6650 XT | llama 7B | Q4_0 | llama.cpp (Vulkan) | 64.53 | 1088.9 | – | – | tg128 | 2025-11-25 | community | – | source |
| GeForce RTX 4060 Ti | llama 7B | Q4_0 | llama.cpp (CUDA) | 64.03 | 3803.45 | – | – | tg128 | 2025-08-02 | community | – | source |
| Radeon RX 9060 XT | llama 7B | Q4_0 | llama.cpp (ROCm) | 63.98 | 2427.16 | – | – | tg128 | 2025-12-15 | community | – | source |
| M3 Ultra (80 GPU) | llama 7B | Q8_0 | llama.cpp (Metal) | 63.93 | 1487.51 | – | – | – | 2023-11-22 | community | – | source |
| M5 Pro | llama 7B | Q4_0 | llama.cpp (Metal) | 63.86 | 1350.43 | – | – | tg128 | 2026-04-14 | community | – | source |
| GeForce RTX 4060 Ti | llama 7B | Q4_0 | llama.cpp (CUDA) | 63.86 | 3394.63 | – | – | tg128 | 2025-08-02 | community | – | source |
| GeForce RTX 5060 Ti | llama 7B | Q4_0 | llama.cpp (Vulkan) | 63.79 | 518.53 | – | – | tg128 | 2026-05-08 | community | contradiction | source |
| M5 Pro | llama 7B | Q4_0 | llama.cpp (Metal) | 63.74 | 1135.71 | – | – | tg128 | 2026-04-14 | community | – | source |
| RTX 3070 | llama 7B | Q4_0 | llama.cpp (Vulkan) | 63.64 | 1689.4 | – | – | tg128 | 2025-10-20 | community | – | source |
| M3 Ultra (60 GPU) | llama 7B | Q8_0 | llama.cpp (Metal) | 63.55 | 1085.76 | – | – | – | 2023-11-22 | community | – | source |
| M2 | llama 7B | Q4_0 | llama.cpp (Vulkan) | 63.16 | 678.14 | – | – | tg128 | 2025-11-26 | community | – | source |
| M2 | llama 7B | Q4_0 | llama.cpp (Vulkan) | 63.1 | 675.02 | – | – | tg128 | 2025-11-28 | community | – | source |
| RTX 3070 | llama 7B | Q4_0 | llama.cpp (Vulkan) | 62.92 | 1832.07 | – | – | tg128 | 2025-10-20 | community | – | source |
| M2 | llama 7B | Q4_0 | llama.cpp (Vulkan) | 62.75 | 53.16 | – | – | tg128 | 2025-11-26 | community | – | source |
| GeForce RTX 3070 | llama 7B | Q4_0 | llama.cpp (CUDA) | 62.17 | 1735.28 | – | – | tg128 | 2025-10-17 | community | – | source |
| RTX 3080 | llama 7B | Q4_0 | llama.cpp (Vulkan) | 62.16 | 1706.07 | – | – | tg128 | 2024-12-18 | community | – | source |
| M2 Ultra (60 GPU) | llama 7B | Q8_0 | llama.cpp (Metal) | 62.14 | 1003.16 | – | – | – | 2023-11-22 | community | – | source |
| Radeon RX 6650 XT | llama 7B | Q4_0 | llama.cpp (Vulkan) | 62.14 | 1029.52 | – | – | tg128 | 2025-11-25 | community | – | source |
| V100 | llama 7B | Q4_0 | llama.cpp (CUDA) | 62.06 | 788.45 | – | – | tg128 | 2025-11-26 | community | – | source |
| RTX 5060 Ti | llama 7B | Q4_0 | llama.cpp (Vulkan) | 62.05 | 2199.15 | – | – | tg128 | 2026-02-03 | community | – | source |
| Radeon RX 5700 XT | llama 7B | Q4_0 | llama.cpp (ROCm) | 62.02 | 314.17 | – | – | tg128 | 2025-01-12 | community | – | source |
| P100 | llama 7B | Q4_0 | llama.cpp (CUDA) | 61.99 | 787.36 | – | – | tg128 | 2025-11-26 | community | – | source |
| NVIDIA RTX 4000 | llama 7B | Q4_0 | llama.cpp (CUDA) | 61.83 | 2779.77 | – | – | tg128 | 2025-09-30 | community | – | source |
| NVIDIA RTX 4000 | llama 7B | Q4_0 | llama.cpp (CUDA) | 61.37 | 3171.86 | – | – | tg128 | 2025-09-30 | community | – | source |
| M1 Max (32 GPU) | llama 7B | Q4_0 | llama.cpp (Metal) | 61.19 | 530.06 | – | – | – | 2023-11-22 | community | – | source |
| Radeon RX 6750 XT | llama 7B | Q4_0 | llama.cpp (Vulkan) | 61.08 | 525.04 | – | – | tg128 | 2026-07-20 | community | – | source |
| M2 Max (30 GPU) | llama 7B | Q4_0 | llama.cpp (Metal) | 60.99 | 537.6 | – | – | – | 2023-11-22 | community | – | source |
| Radeon RX 7600 | llama 7B | Q4_0 | llama.cpp (Vulkan) | 60.88 | 1353.47 | – | – | tg128 | 2026-07-30 | community | – | source |
| GB10 | llama 7B | Q4_0 | llama.cpp (Vulkan) | 60.39 | 3321.91 | – | – | tg128 | 2026-07-14 | community | – | source |
| GB10 | llama 7B | Q4_0 | llama.cpp (CUDA) | 60.24 | 3310.48 | – | – | tg128 | 2026-01-20 | community | – | source |
| Radeon RX 7900 XT | llama 7B | Q4_0 | llama.cpp (Vulkan) | 60.18 | 721.44 | – | – | tg128 | 2025-09-24 | community | – | source |
| RTX 5060 Ti | llama 7B | Q4_0 | llama.cpp (Vulkan) | 60.02 | 689.59 | – | – | tg128 | 2026-02-03 | community | – | source |
| RTX 4060 | llama 7B | Q4_0 | llama.cpp (Vulkan) | 60.01 | 2358.03 | – | – | tg128 | 2025-11-04 | community | – | source |
| M1 Ultra (64 GPU) | llama 7B | Q8_0 | llama.cpp (Metal) | 59.87 | 1042.95 | – | – | – | 2023-11-22 | community | – | source |
| GeForce RTX 3070 | llama 7B | Q4_0 | llama.cpp (CUDA) | 59.72 | 1729.14 | – | – | tg128 | 2025-10-17 | community | – | source |
| GB10 | llama 7B | Q4_0 | llama.cpp (CUDA) | 59.62 | 4016.46 | – | – | tg128 | 2026-03-30 | community | – | source |
| GeForce RTX 4060 | llama 7B | Q4_0 | llama.cpp (CUDA) | 59.55 | 2676.6 | – | – | tg128 | 2026-02-11 | community | – | source |
| RTX 4060 | llama 7B | Q4_0 | llama.cpp (Vulkan) | 59.53 | 2135.66 | – | – | tg128 | 2025-11-04 | community | – | source |
| GeForce RTX 3070 | llama 7B | Q4_0 | llama.cpp (CUDA) | 59.52 | 1918.79 | – | – | tg128 | 2025-10-17 | community | – | source |
| Radeon RX 6800 XT | llama 7B | Q4_0 | llama.cpp (Vulkan) | 59.36 | 488.06 | – | – | tg128 | 2025-09-01 | community | – | source |
| Radeon RX 6650 XT | llama 7B | Q4_0 | llama.cpp (Vulkan) | 59.22 | 735.64 | – | – | tg128 | 2025-06-10 | community | – | source |
| Radeon RX 6800 XT | llama 7B | Q4_0 | llama.cpp (Vulkan) | 59.22 | 484.37 | – | – | tg128 | 2025-09-01 | community | – | source |
| Radeon RX 6650 XT | llama 7B | Q4_0 | llama.cpp (Vulkan) | 59.18 | 730.64 | – | – | tg128 | 2025-06-10 | community | – | source |
| GeForce RTX 4060 | llama 7B | Q4_0 | llama.cpp (CUDA) | 59.15 | 2928.74 | – | – | tg128 | 2026-02-11 | community | – | source |
| GB10 | llama 7B | Q4_0 | llama.cpp (CUDA) | 59.06 | 3306.61 | – | – | tg128 | 2026-03-30 | community | – | source |
| GB10 | llama 7B | Q4_0 | llama.cpp (CUDA) | 59.05 | 3954.97 | – | – | tg128 | 2026-03-30 | community | – | source |
| Radeon RX 7600 XT | llama 7B | Q4_0 | llama.cpp (Vulkan) | 59.02 | 586.16 | – | – | tg128 | 2025-05-10 | community | – | source |
| Radeon RX 7600 | llama 7B | Q4_0 | llama.cpp (ROCm) | 58.73 | 1308.31 | – | – | tg128 | 2026-07-14 | community | – | source |
| GB10 | llama 7B | Q4_0 | llama.cpp (Vulkan) | 58.51 | 2789.51 | – | – | tg128 | 2026-07-14 | community | – | source |
| M2 | llama 7B | Q4_0 | llama.cpp (Vulkan) | 58.47 | 53.1 | – | – | tg128 | 2025-11-26 | community | – | source |
| Radeon RX 7600 XT | llama 7B | Q4_0 | llama.cpp (Vulkan) | 58.44 | 632.88 | – | – | tg128 | 2025-05-10 | community | – | source |
| GB10 | llama 7B | Q4_0 | llama.cpp (CUDA) | 58.38 | 3261.99 | – | – | tg128 | 2026-03-30 | community | – | source |
| P100 | llama 7B | Q4_0 | llama.cpp (CUDA) | 58.35 | 760.8 | – | – | tg128 | 2025-11-26 | community | – | source |
| V100 | llama 7B | Q4_0 | llama.cpp (CUDA) | 58.33 | 767.12 | – | – | tg128 | 2025-11-26 | community | – | source |
| GeForce RTX 4060 | llama 7B | Q4_0 | llama.cpp (CUDA) | 58.22 | 2466.31 | – | – | tg128 | 2026-02-11 | community | – | source |
| GB10 | llama 7B | Q4_0 | llama.cpp (CUDA) | 58.14 | 3861.19 | – | – | tg128 | 2026-02-16 | community | – | source |
| GB10 | llama 7B | Q4_0 | llama.cpp (CUDA) | 58.13 | 3241.51 | – | – | tg128 | 2026-02-16 | community | – | source |
| Radeon RX 7600 | llama 7B | Q4_0 | llama.cpp (Vulkan) | 58.03 | 1167.2 | – | – | tg128 | 2026-07-30 | community | – | source |
| Radeon RX 7600 | llama 7B | Q4_0 | llama.cpp (ROCm) | 58.0 | 1312.19 | – | – | tg128 | 2026-07-30 | community | – | source |
| GeForce RTX 4060 | llama 7B | Q4_0 | llama.cpp (CUDA) | 57.87 | 2464.58 | – | – | tg128 | 2026-02-11 | community | – | source |
| GeForce RTX 3070 | llama 7B | Q4_0 | llama.cpp (CUDA) | 57.36 | 1904.62 | – | – | tg128 | 2025-10-17 | community | – | source |
| GB10 | llama 7B | Q4_0 | llama.cpp (CUDA) | 57.21 | 3062.31 | – | – | tg128 | 2025-10-15 | community | – | source |
| GB10 | llama 7B | Q4_0 | llama.cpp (CUDA) | 56.74 | 3661.37 | – | – | tg128 | 2025-10-15 | community | – | source |
| M3 Max (30 GPU) | llama 7B | Q4_0 | llama.cpp (Metal) | 56.58 | 567.59 | – | – | – | 2023-11-22 | community | – | source |
| Radeon RX 7900 XTX | llama 7B | Q4_0 | llama.cpp (Vulkan) | 56.35 | 1727.1 | – | – | tg128 | 2026-08-23 | community | – | source |
| Radeon RX 7600 XT | llama 7B | Q4_0 | llama.cpp (Vulkan) | 56.11 | 1024.38 | – | – | tg128 | 2026-02-15 | community | – | source |
| M1 Ultra (48 GPU) | llama 7B | Q8_0 | llama.cpp (Metal) | 55.69 | 783.45 | – | – | – | 2023-11-22 | community | – | source |
| Radeon RX 7900 XT | llama 7B | Q4_0 | llama.cpp (Vulkan) | 55.41 | 722.38 | – | – | tg128 | 2025-09-24 | community | – | source |
| Radeon 8060S | llama 7B | Q4_0 | llama.cpp (Vulkan) | 55.32 | 1147.07 | – | – | tg128 | 2025-12-11 | community | – | source |
| Radeon RX 7600 XT | llama 7B | Q4_0 | llama.cpp (Vulkan) | 55.2 | 1028.16 | – | – | tg128 | 2026-02-15 | community | – | source |
| Radeon RX 9060 XT | llama 7B | Q4_0 | llama.cpp (Vulkan) | 54.96 | 1651.6 | – | – | tg128 | 2025-12-30 | community | – | source |
| GB10 | llama 7B | Q4_0 | llama.cpp (CUDA) | 54.91 | 3920.63 | – | – | tg128 | 2026-03-30 | community | – | source |
| Radeon RX 9060 XT | llama 7B | Q4_0 | llama.cpp (Vulkan) | 54.84 | 418.46 | – | – | tg128 | 2025-12-30 | community | – | source |
| Radeon 8060S | llama 7B | Q4_0 | llama.cpp (Vulkan) | 54.76 | 1135.17 | – | – | tg128 | 2025-11-19 | community | – | source |
| M1 Max (24 GPU) | llama 7B | Q4_0 | llama.cpp (Metal) | 54.61 | 400.26 | – | – | – | 2023-11-22 | community | – | source |
| GB10 | llama 7B | Q4_0 | llama.cpp (CUDA) | 54.39 | 3293.78 | – | – | tg128 | 2026-03-30 | community | – | source |
| Arc A770 | llama 7B | Q4_0 | llama.cpp (SYCL) | 54.38 | 979.76 | – | – | tg128 | 2026-09-17 | community | – | source |
| Radeon 8060S | llama 7B | Q4_0 | llama.cpp (Vulkan) | 54.38 | 1704.3 | – | – | tg128 | 2026-10-02 | community | – | source |
| RTX 3060 | llama 7B | Q4_0 | llama.cpp (Vulkan) | 54.28 | 1298.03 | – | – | tg128 | 2025-01-21 | community | – | source |
| M4 Max (40 GPU) | llama 7B | Q8_0 | llama.cpp (Metal) | 54.05 | 891.94 | – | – | – | 2023-11-22 | community | – | source |
| GeForce RTX 5060 Ti | llama 7B | Q4_0 | llama.cpp (CUDA) | 53.96 | 1037.02 | – | – | tg128 | 2026-05-08 | community | contradiction | source |
| Radeon RX 6600 XT | llama 7B | Q4_0 | llama.cpp (Vulkan) | 53.92 | 574.65 | – | – | tg128 | 2025-01-14 | community | – | source |
| RTX 4060 | llama 7B | Q4_0 | llama.cpp (Vulkan) | 53.8 | 1908.24 | – | – | tg128 | 2026-10-10 | community | – | source |
| Radeon RX 7600 XT | llama 7B | Q4_0 | llama.cpp (Vulkan) | 53.7 | 612.33 | – | – | tg128 | 2025-08-01 | community | – | source |
| GeForce RTX 5060 Ti | llama 7B | Q4_0 | llama.cpp (CUDA) | 53.65 | 428.35 | – | – | tg128 | 2026-05-08 | community | contradiction | source |
| GB10 | llama 7B | Q4_0 | llama.cpp (Vulkan) | 53.64 | 3279.89 | – | – | tg128 | 2026-03-15 | community | – | source |
| Radeon 8060S | llama 7B | Q4_0 | llama.cpp (Vulkan) | 53.59 | 1288.96 | – | – | tg128 | 2025-09-21 | community | – | source |
| Arc Pro B60 | llama 7B | Q4_0 | llama.cpp (Vulkan) | 53.56 | 1581.29 | – | – | tg128 | 2026-06-04 | community | contradiction | source |
| Radeon RX 7600 | llama 7B | Q4_0 | llama.cpp (ROCm) | 53.48 | 1076.93 | – | – | tg128 | 2026-07-30 | community | – | source |
| Radeon 8060S | llama 7B | Q4_0 | llama.cpp (Vulkan) | 53.37 | 1083.51 | – | – | tg128 | 2025-11-12 | community | – | source |
| Radeon RX 7600 | llama 7B | Q4_0 | llama.cpp (ROCm) | 53.27 | 1006.77 | – | – | tg128 | 2026-07-14 | community | – | source |
| Radeon RX 6600 | llama 7B | Q4_0 | llama.cpp (Vulkan) | 53.24 | 808.76 | – | – | tg128 | 2026-03-21 | community | – | source |
| GB10 | llama 7B | Q4_0 | llama.cpp (Vulkan) | 53.21 | 3219.57 | – | – | tg128 | 2026-03-15 | community | – | source |
| Radeon 8060S | llama 7B | Q4_0 | llama.cpp (Vulkan) | 53.21 | 1120.83 | – | – | tg128 | 2025-10-24 | community | – | source |
| Radeon RX 6600 | llama 7B | Q4_0 | llama.cpp (Vulkan) | 53.17 | 804.64 | – | – | tg128 | 2026-07-18 | community | – | source |
| A770 | llama 7B | Q4_0 | llama.cpp (Vulkan) | 53.07 | 1119.68 | – | – | tg128 | 2026-03-24 | community | – | source |
| Radeon RX 7600 XT | llama 7B | Q4_0 | llama.cpp (Vulkan) | 53.02 | 840.85 | – | – | tg128 | 2026-02-15 | community | – | source |
| Radeon 8060S | llama 7B | Q4_0 | llama.cpp (Vulkan) | 53.0 | 1357.07 | – | – | tg128 | 2025-09-21 | community | – | source |
| A770 | llama 7B | Q4_0 | llama.cpp (SYCL) | 52.98 | 2173.52 | – | – | tg128 | 2025-07-30 | community | – | source |
| GeForce RTX 5060 Ti | llama 7B | Q4_0 | llama.cpp (CUDA) | 52.85 | 416.06 | – | – | tg128 | 2026-05-08 | community | contradiction | source |
| Radeon RX 7600 XT | llama 7B | Q4_0 | llama.cpp (Vulkan) | 52.84 | 606.24 | – | – | tg128 | 2025-08-01 | community | – | source |
| Radeon RX 7600 XT | llama 7B | Q4_0 | llama.cpp (Vulkan) | 52.8 | 544.83 | – | – | tg128 | 2026-02-15 | community | – | source |
| A770 | llama 7B | Q4_0 | llama.cpp (SYCL) | 52.76 | 916.88 | – | – | tg128 | 2026-01-28 | community | contradiction | source |
| Radeon RX 7600 XT | llama 7B | Q4_0 | llama.cpp (Vulkan) | 52.69 | 1039.52 | – | – | tg128 | 2026-02-15 | community | – | source |
| Radeon 8060S | llama 7B | Q4_0 | llama.cpp (Vulkan) | 52.69 | 1243.31 | – | – | tg128 | 2025-10-24 | community | – | source |
| RTX 3060 | llama 7B | Q4_0 | llama.cpp (Vulkan) | 52.66 | 428.54 | – | – | tg128 | 2025-11-26 | community | – | source |
| A770 | llama 7B | Q4_0 | llama.cpp (SYCL) | 52.62 | 1866.14 | – | – | tg128 | 2025-09-21 | community | contradiction | source |
| RTX 4060 | llama 7B | Q4_0 | llama.cpp (Vulkan) | 52.58 | 1793.16 | – | – | tg128 | 2026-10-10 | community | – | source |
| A770 | llama 7B | Q4_0 | llama.cpp (Vulkan) | 52.56 | 1073.85 | – | – | tg128 | 2026-03-24 | community | – | source |
| Radeon 8060S | llama 7B | Q4_0 | llama.cpp (Vulkan) | 52.5 | 1264.6 | – | – | tg128 | 2025-10-24 | community | – | source |
| Arc A770 | llama 7B | Q4_0 | llama.cpp (Vulkan) | 52.48 | 296.62 | – | – | tg128 | 2025-11-04 | community | – | source |
| Radeon RX 6600 | llama 7B | Q4_0 | llama.cpp (Vulkan) | 52.47 | 410.19 | – | – | tg128 | 2025-12-21 | community | – | source |
| GB10 | llama 7B | Q4_0 | llama.cpp (Vulkan) | 52.28 | 2737.79 | – | – | tg128 | 2026-03-15 | community | – | source |
| Radeon 8060S | llama 7B | Q4_0 | llama.cpp (Vulkan) | 52.28 | 972.06 | – | – | tg128 | 2025-10-23 | community | – | source |
| A770 | llama 7B | Q4_0 | llama.cpp (SYCL) | 52.26 | 2185.36 | – | – | tg128 | 2025-07-30 | community | – | source |
| P100 | llama 7B | Q4_0 | llama.cpp (CUDA) | 52.24 | 1076.52 | – | – | tg128 | 2025-11-27 | community | – | source |
| Radeon 8060S | llama 7B | Q4_0 | llama.cpp (Vulkan) | 52.23 | 1044.51 | – | – | tg128 | 2025-12-11 | community | – | source |
| A770 | llama 7B | Q4_0 | llama.cpp (Vulkan) | 52.04 | 986.03 | – | – | tg128 | 2025-09-21 | community | – | source |
| Radeon 8060S | llama 7B | Q4_0 | llama.cpp (Vulkan) | 52.03 | 1305.74 | – | – | tg128 | 2025-10-24 | community | – | source |
| Radeon 8060S | llama 7B | Q4_0 | llama.cpp (Vulkan) | 52.01 | 869.4 | – | – | tg128 | 2025-09-10 | community | – | source |
| A2000 | llama 7B | Q4_0 | llama.cpp (CUDA) | 51.96 | 1411.42 | – | – | tg128 | 2025-08-23 | community | – | source |
| Radeon 8060S | llama 7B | Q4_0 | llama.cpp (Vulkan) | 51.89 | 803.41 | – | – | tg128 | 2025-08-17 | community | – | source |
| Radeon 8060S | llama 7B | Q4_0 | llama.cpp (Vulkan) | 51.81 | 1487.08 | – | – | tg128 | 2026-10-02 | community | – | source |
| Radeon 8060S | llama 7B | Q4_0 | llama.cpp (Vulkan) | 51.71 | 951.25 | – | – | tg128 | 2025-09-10 | community | – | source |
| GB10 | llama 7B | Q4_0 | llama.cpp (Vulkan) | 51.7 | 2710.58 | – | – | tg128 | 2026-03-15 | community | – | source |
| Radeon 8060S | llama 7B | Q4_0 | llama.cpp (Vulkan) | 51.68 | 1041.42 | – | – | tg128 | 2025-11-19 | community | – | source |
| Radeon 8060S | llama 7B | Q4_0 | llama.cpp (Vulkan) | 51.67 | 1023.92 | – | – | tg128 | 2025-11-12 | community | – | source |
| Radeon 8060S | llama 7B | Q4_0 | llama.cpp (Vulkan) | 51.62 | 1054.53 | – | – | tg128 | 2025-10-23 | community | – | source |
| P100 | llama 7B | Q4_0 | llama.cpp (CUDA) | 51.54 | 1006.05 | – | – | tg128 | 2025-11-27 | community | – | source |
| Arc A770 | llama 7B | Q4_0 | llama.cpp (Vulkan) | 51.43 | 1185.04 | – | – | tg128 | 2025-11-04 | community | – | source |
| Radeon RX 7600 XT | llama 7B | Q4_0 | llama.cpp (Vulkan) | 51.42 | 494.8 | – | – | tg128 | 2025-05-06 | community | – | source |
| Radeon RX 6800M | llama 7B | Q4_0 | llama.cpp (Vulkan) | 51.36 | 442.71 | – | – | tg128 | 2025-07-22 | community | – | source |
| Arc A770 | llama 7B | Q4_0 | llama.cpp (Vulkan) | 51.36 | 251.11 | – | – | tg128 | 2025-11-04 | community | – | source |
| Radeon 8060S | llama 7B | Q4_0 | llama.cpp (ROCm) | 51.33 | 1456.15 | – | – | tg128 | 2026-10-02 | community | – | source |
| GeForce RTX 5060 Ti | llama 7B | Q4_0 | llama.cpp (CUDA) | 51.3 | 913.16 | – | – | tg128 | 2026-05-08 | community | contradiction | source |
| A2000 | llama 7B | Q4_0 | llama.cpp (CUDA) | 51.21 | 1259.66 | – | – | tg128 | 2025-08-23 | community | – | source |
| Radeon 8060S | llama 7B | Q4_0 | llama.cpp (Vulkan) | 51.14 | 927.96 | – | – | tg128 | 2025-08-17 | community | – | source |
| P100 | llama 7B | Q4_0 | llama.cpp (CUDA) | 51.08 | 735.19 | – | – | tg128 | 2025-08-23 | community | – | source |
| A770 | llama 7B | Q4_0 | llama.cpp (Vulkan) | 50.9 | 903.49 | – | – | tg128 | 2025-09-21 | community | – | source |
| GB10 | llama 7B | Q4_0 | llama.cpp (Vulkan) | 50.8 | 2665.61 | – | – | tg128 | 2026-01-20 | community | – | source |
| Radeon 8060S | llama 7B | Q4_0 | llama.cpp (ROCm) | 50.75 | 1117.55 | – | – | tg128 | 2025-11-15 | community | – | source |
| M4 Pro (20 GPU) | llama 7B | Q4_0 | llama.cpp (Metal) | 50.74 | 439.78 | – | – | – | 2023-11-22 | community | – | source |
| Radeon 8060S | llama 7B | Q4_0 | llama.cpp (ROCm) | 50.69 | 1118.89 | – | – | tg128 | 2025-11-15 | community | – | source |
| P100 | llama 7B | Q4_0 | llama.cpp (CUDA) | 50.66 | 492.2 | – | – | tg128 | 2025-11-27 | community | – | source |
| Radeon RX 6600 | llama 7B | Q4_0 | llama.cpp (Vulkan) | 50.63 | 761.89 | – | – | tg128 | 2026-03-21 | community | – | source |
| Radeon RX 7600 XT | llama 7B | Q4_0 | llama.cpp (Vulkan) | 50.63 | 483.22 | – | – | tg128 | 2025-05-21 | community | – | source |
| Radeon 8060S | llama 7B | Q4_0 | llama.cpp (ROCm) | 50.59 | 1012.79 | – | – | tg128 | 2025-11-26 | community | – | source |
| Radeon RX 6600 | llama 7B | Q4_0 | llama.cpp (Vulkan) | 50.57 | 758.5 | – | – | tg128 | 2026-07-18 | community | – | source |
| Radeon 8060S | llama 7B | Q4_0 | llama.cpp (ROCm) | 50.57 | 1576.59 | – | – | tg128 | 2025-12-11 | community | – | source |
| Arc B570 | llama 7B | Q4_0 | llama.cpp (Vulkan) | 50.49 | 288.51 | – | – | tg128 | 2025-09-21 | community | – | source |
| Arc A770 | llama 7B | Q4_0 | llama.cpp (Vulkan) | 50.47 | 926.02 | – | – | tg128 | 2025-11-04 | community | – | source |
| Radeon 8060S | llama 7B | Q4_0 | llama.cpp (ROCm) | 50.4 | 1488.56 | – | – | tg128 | 2026-02-25 | community | – | source |
| Arc A770 | llama 7B | Q4_0 | llama.cpp (Vulkan) | 50.39 | 244.24 | – | – | tg128 | 2025-09-01 | community | – | source |
| Radeon 8060S | llama 7B | Q4_0 | llama.cpp (ROCm) | 50.3 | 998.73 | – | – | tg128 | 2025-12-20 | community | – | source |
| Radeon 8060S | llama 7B | Q4_0 | llama.cpp (ROCm) | 50.29 | 1464.06 | – | – | tg128 | 2025-12-15 | community | – | source |
| RTX 3060 | llama 7B | Q4_0 | llama.cpp (Vulkan) | 50.23 | 439.87 | – | – | tg128 | 2025-11-26 | community | – | source |
| P100 | llama 7B | Q4_0 | llama.cpp (CUDA) | 50.2 | 703.27 | – | – | tg128 | 2025-08-23 | community | – | source |
| P100 | llama 7B | Q4_0 | llama.cpp (CUDA) | 50.17 | 476.77 | – | – | tg128 | 2025-11-27 | community | – | source |
| Radeon RX 6600 | llama 7B | Q4_0 | llama.cpp (Vulkan) | 50.11 | 404.04 | – | – | tg128 | 2025-12-21 | community | – | source |
| Radeon RX 6600 | llama 7B | Q4_0 | llama.cpp (Vulkan) | 49.76 | 331.58 | – | – | tg128 | 2025-03-10 | community | – | source |
| A770 | llama 7B | Q4_0 | llama.cpp (Vulkan) | 49.66 | 372.22 | – | – | tg128 | 2025-09-21 | community | – | source |
| M4 Pro (16 GPU) | llama 7B | Q4_0 | llama.cpp (Metal) | 49.64 | 364.06 | – | – | – | 2023-11-22 | community | – | source |
| Arc B570 | llama 7B | Q4_0 | llama.cpp (Vulkan) | 49.64 | 913.95 | – | – | tg128 | 2025-09-21 | community | – | source |
| Radeon RX 7600S | llama 7B | Q4_0 | llama.cpp (ROCm) | 49.62 | 915.55 | – | – | tg128 | 2026-05-09 | community | – | source |
| P102-100 | llama 7B | Q4_0 | llama.cpp (CUDA) | 49.59 | 1032.13 | – | – | tg128 | 2026-10-01 | community | – | source |
| Radeon RX 6800M | llama 7B | Q4_0 | llama.cpp (Vulkan) | 49.58 | 663.96 | – | – | tg128 | 2025-07-22 | community | – | source |
| Arc A770 | llama 7B | Q4_0 | llama.cpp (Vulkan) | 49.55 | 762.02 | – | – | tg128 | 2025-09-01 | community | – | source |
| Radeon 8060S | llama 7B | Q4_0 | llama.cpp (ROCm) | 49.5 | 1223.09 | – | – | tg128 | 2026-04-03 | community | – | source |
| RTX 5060 Ti | llama 7B | Q4_0 | llama.cpp (Vulkan) | 49.44 | 2021.05 | – | – | tg128 | 2026-02-03 | community | – | source |
| A770 | llama 7B | Q4_0 | llama.cpp (Vulkan) | 49.43 | 725.31 | – | – | tg128 | 2025-05-26 | community | – | source |
| Arc A770 | llama 7B | Q4_0 | llama.cpp (Vulkan) | 49.34 | 295.72 | – | – | tg128 | 2025-09-21 | community | – | source |
| Radeon 8060S | llama 7B | Q4_0 | llama.cpp (ROCm) | 49.1 | 1143.51 | – | – | tg128 | 2026-08-11 | community | – | source |
| P102-100 | llama 7B | Q4_0 | llama.cpp (CUDA) | 49.08 | 1029.43 | – | – | tg128 | 2026-10-01 | community | – | source |
| Radeon RX 6800M | llama 7B | Q4_0 | llama.cpp (Vulkan) | 49.06 | 784.16 | – | – | tg128 | 2025-07-17 | community | – | source |
| Radeon RX 6800M | llama 7B | Q4_0 | llama.cpp (Vulkan) | 49.06 | 784.16 | – | – | tg128 | 2025-07-22 | community | – | source |
| RTX 3060 | llama 7B | Q4_0 | llama.cpp (Vulkan) | 49.03 | 1059.76 | – | – | tg128 | 2025-03-20 | community | – | source |
| A770 | llama 7B | Q4_0 | llama.cpp (Vulkan) | 49.02 | 603.46 | – | – | tg128 | 2025-04-03 | community | – | source |
| Radeon 8060S | llama 7B | Q4_0 | llama.cpp (ROCm) | 48.97 | 366.08 | – | – | tg128 | 2025-09-23 | community | – | source |
| Radeon RX 6800M | llama 7B | Q4_0 | llama.cpp (Vulkan) | 48.71 | 861.99 | – | – | tg128 | 2025-07-17 | community | – | source |
| Radeon RX 6800M | llama 7B | Q4_0 | llama.cpp (Vulkan) | 48.71 | 861.99 | – | – | tg128 | 2025-07-22 | community | – | source |
| Radeon RX 480 G | llama 7B | Q4_0 | llama.cpp (Vulkan) | 48.64 | 325.3 | – | – | tg128 | 2025-11-26 | community | – | source |
| Radeon RX 7600 XT | llama 7B | Q4_0 | llama.cpp (ROCm) | 48.58 | 1099.64 | – | – | tg128 | 2025-08-01 | community | – | source |
| Radeon RX 6600 | llama 7B | Q4_0 | llama.cpp (Vulkan) | 48.52 | 617.85 | – | – | tg128 | 2025-08-15 | community | – | source |
| Arc A770 | llama 7B | Q4_0 | llama.cpp (Vulkan) | 48.46 | 904.11 | – | – | tg128 | 2025-09-21 | community | – | source |
| A770 | llama 7B | Q4_0 | llama.cpp (Vulkan) | 48.45 | 1116.8 | – | – | tg128 | 2025-04-01 | community | – | source |
| GB10 | llama 7B | Q4_0 | llama.cpp (Vulkan) | 48.42 | 2692.06 | – | – | tg128 | 2026-03-15 | community | – | source |
| Radeon RX 7600 XT | llama 7B | Q4_0 | llama.cpp (Vulkan) | 48.31 | 420.83 | – | – | tg128 | 2025-05-21 | community | – | source |
| Radeon RX 6600 | llama 7B | Q4_0 | llama.cpp (Vulkan) | 48.31 | 622.72 | – | – | tg128 | 2025-08-15 | community | – | source |
| A770 | llama 7B | Q4_0 | llama.cpp (Vulkan) | 48.3 | 363.82 | – | – | tg128 | 2025-09-21 | community | – | source |
| Radeon RX 6600M | llama 7B | Q4_0 | llama.cpp (Vulkan) | 48.21 | 605.59 | – | – | tg128 | 2025-04-10 | community | – | source |
| A770 | llama 7B | Q4_0 | llama.cpp (Vulkan) | 48.17 | 327.58 | – | – | tg128 | 2025-05-26 | community | – | source |
| Radeon 8060S | llama 7B | Q4_0 | llama.cpp (ROCm) | 47.97 | 351.36 | – | – | tg128 | 2025-09-23 | community | – | source |
| Radeon RX 6600 XT | llama 7B | Q4_0 | llama.cpp (Vulkan) | 47.94 | 428.92 | – | – | tg128 | 2025-02-08 | community | – | source |
| Radeon RX 7600 XT | llama 7B | Q4_0 | llama.cpp (Vulkan) | 47.94 | 717.53 | – | – | tg128 | 2025-05-21 | community | – | source |
| M1 Ultra | llama 7B | F16 | llama.cpp (Metal) | 47.82 | 1294.31 | – | – | tg128 | 2026-08-28 | community | – | source |
| Arc A770 | llama 7B | Q4_0 | llama.cpp (SYCL) | 47.81 | 826.15 | – | – | tg128 | 2026-09-17 | community | – | source |
| Arc Pro B50 | llama 7B | Q4_0 | llama.cpp (SYCL) | 47.8 | 397.73 | – | – | tg128 | 2026-05-28 | community | – | source |
| Arc Pro B50 | llama 7B | Q4_0 | llama.cpp (Vulkan) | 47.69 | 1122.53 | – | – | tg128 | 2026-05-30 | community | – | source |
| Radeon RX 7600 XT | llama 7B | Q4_0 | llama.cpp (ROCm) | 47.65 | 1199.16 | – | – | tg128 | 2025-08-01 | community | – | source |
| GB10 | llama 7B | Q4_0 | llama.cpp (Vulkan) | 47.55 | 2731.93 | – | – | tg128 | 2026-03-14 | community | – | source |
| Radeon RX 6600 | llama 7B | Q4_0 | llama.cpp (Vulkan) | 47.49 | 615.03 | – | – | tg128 | 2025-06-27 | community | – | source |
| Radeon RX 6600 | llama 7B | Q4_0 | llama.cpp (Vulkan) | 47.47 | 380.87 | – | – | tg128 | 2025-03-10 | community | – | source |
| Radeon 8060S | llama 7B | Q4_0 | llama.cpp (ROCm) | 47.36 | 1313.76 | – | – | tg128 | 2026-10-02 | community | – | source |
| RTX 5060 Ti | llama 7B | Q4_0 | llama.cpp (Vulkan) | 47.35 | 511.59 | – | – | tg128 | 2026-02-03 | community | – | source |
| Radeon RX 7600 XT | llama 7B | Q4_0 | llama.cpp (Vulkan) | 46.98 | 694.27 | – | – | tg128 | 2025-05-21 | community | – | source |
| P102-100 | llama 7B | Q4_0 | llama.cpp (CUDA) | 46.77 | 935.02 | – | – | tg128 | 2026-10-01 | community | – | source |
| Radeon 8060S | llama 7B | Q4_0 | llama.cpp (Vulkan) | 46.69 | 1121.72 | – | – | tg128 | 2025-08-17 | community | contradiction | source |
| Radeon 8060S | llama 7B | Q4_0 | llama.cpp (ROCm) | 46.68 | 1009.02 | – | – | tg128 | 2025-11-15 | community | – | source |
| GB10 | llama 7B | Q4_0 | llama.cpp (Vulkan) | 46.67 | 3137.97 | – | – | tg128 | 2026-03-15 | community | – | source |
| Radeon 8060S | llama 7B | Q4_0 | llama.cpp (ROCm) | 46.65 | 1004.84 | – | – | tg128 | 2025-11-15 | community | – | source |
| GB10 | llama 7B | Q4_0 | llama.cpp (Vulkan) | 46.61 | 3171.16 | – | – | tg128 | 2026-03-14 | community | – | source |
| Radeon 8060S | llama 7B | Q4_0 | llama.cpp (ROCm) | 46.51 | 908.35 | – | – | tg128 | 2025-11-26 | community | – | source |
| Radeon 8060S | llama 7B | Q4_0 | llama.cpp (ROCm) | 46.46 | 1352.52 | – | – | tg128 | 2025-12-11 | community | – | source |
| P102-100 | llama 7B | Q4_0 | llama.cpp (CUDA) | 46.45 | 927.54 | – | – | tg128 | 2026-10-01 | community | – | source |
| T4 | llama 7B | Q4_0 | llama.cpp (CUDA) | 46.38 | 1219.06 | – | – | tg128 | 2025-08-14 | community | – | source |
| Arc Pro B50 | llama 7B | Q4_0 | llama.cpp (Vulkan) | 46.3 | 1186.45 | – | – | tg128 | 2026-05-30 | community | – | source |
| Radeon RX 480 G | llama 7B | Q4_0 | llama.cpp (Vulkan) | 46.18 | 311.9 | – | – | tg128 | 2025-11-26 | community | – | source |
| Radeon 8060S | llama 7B | Q4_0 | llama.cpp (ROCm) | 46.15 | 843.04 | – | – | tg1024 | 2026-04-03 | community | – | source |
| Radeon 8060S | llama 7B | Q4_0 | llama.cpp (Vulkan) | 46.09 | 1347.94 | – | – | tg128 | 2025-08-17 | community | – | source |
| Radeon 8060S | llama 7B | Q4_0 | llama.cpp (ROCm) | 46.01 | 1201.51 | – | – | tg128 | 2026-02-25 | community | – | source |
| Arc Pro B50 | llama 7B | Q4_0 | llama.cpp (SYCL) | 45.92 | 424.5 | – | – | tg128 | 2026-05-28 | community | – | source |
| Arc Pro B50 | llama 7B | Q4_0 | llama.cpp (SYCL) | 45.74 | 397.09 | – | – | tg128 | 2026-05-28 | community | – | source |
| A2000 | llama 7B | Q4_0 | llama.cpp (CUDA) | 45.69 | 1361.85 | – | – | tg128 | 2025-08-23 | community | – | source |
| Radeon 8060S | llama 7B | Q4_0 | llama.cpp (ROCm) | 45.53 | 1168.63 | – | – | tg128 | 2025-12-15 | community | – | source |
| A2000 | llama 7B | Q4_0 | llama.cpp (CUDA) | 45.52 | 1245.19 | – | – | tg128 | 2025-08-23 | community | – | source |
| A770 | llama 7B | Q4_0 | llama.cpp (Vulkan) | 45.22 | 314.24 | – | – | tg128 | 2025-01-10 | community | – | source |
| Radeon RX 7600S | llama 7B | Q4_0 | llama.cpp (ROCm) | 45.02 | 783.23 | – | – | tg128 | 2026-05-09 | community | – | source |
| M4 | llama 7B | Q8_0 | llama.cpp (Metal) | 44.89 | 702.39 | – | – | tg128 | 2025-05-20 | community | – | source |
| Arc A770 | llama 7B | Q4_0 | llama.cpp (Vulkan) | 44.49 | 95.52 | – | – | tg128 | 2025-01-08 | community | – | source |
| Radeon 8060S | llama 7B | Q4_0 | llama.cpp (ROCm) | 44.15 | 1106.09 | – | – | tg128 | 2026-04-03 | community | – | source |
| Radeon RX 9060 XT | llama 7B | Q4_0 | llama.cpp (Vulkan) | 44.1 | 658.95 | – | – | tg128 | 2025-06-13 | community | – | source |
| T4 | llama 7B | Q4_0 | llama.cpp (CUDA) | 44.03 | 1309.73 | – | – | tg128 | 2025-08-14 | community | – | source |
| Arc A750 | llama 7B | Q4_0 | llama.cpp (Vulkan) | 43.96 | 303.37 | – | – | tg128 | 2025-11-07 | community | – | source |
| Arc Pro B60 | llama 7B | Q4_0 | llama.cpp (SYCL) | 43.93 | 399.3 | – | – | tg128 | 2026-01-17 | community | – | source |
| M4 Max (32 GPU) | llama 7B | Q8_0 | llama.cpp (Metal) | 43.87 | 718.56 | – | – | – | 2023-11-22 | community | – | source |
| RTX 5060 Ti | llama 7B | Q4_0 | llama.cpp (Vulkan) | 43.79 | 2784.46 | – | – | tg128 | 2025-07-30 | community | – | source |
| Radeon 8060S | llama 7B | Q4_0 | llama.cpp (ROCm) | 43.79 | 1063.6 | – | – | tg128 | 2026-08-11 | community | – | source |
| RTX 4050 | llama 7B | Q4_0 | llama.cpp (CUDA) | 43.72 | 1725.85 | – | – | tg128 | 2025-01-13 | community | – | source |
| Arc B570 | llama 7B | Q4_0 | llama.cpp (Vulkan) | 43.38 | 285.36 | – | – | tg128 | 2025-08-17 | community | – | source |
| Radeon RX 470 G | llama 7B | Q4_0 | llama.cpp (Vulkan) | 43.35 | 211.25 | – | – | tg128 | 2026-07-09 | community | – | source |
| Radeon RX 480 G | llama 7B | Q4_0 | llama.cpp (Vulkan) | 43.33 | 330.84 | – | – | tg128 | 2025-11-26 | community | – | source |
| Arc A770 | llama 7B | Q4_0 | llama.cpp (SYCL) | 43.18 | 822.07 | – | – | tg128 | 2026-08-31 | community | – | source |
| M3 Max (40 GPU) | llama 7B | Q8_0 | llama.cpp (Metal) | 42.75 | 757.64 | – | – | – | 2023-11-22 | community | – | source |
| Arc B570 | llama 7B | Q4_0 | llama.cpp (Vulkan) | 42.7 | 469.46 | – | – | tg128 | 2025-08-17 | community | – | source |
| Arc A750 | llama 7B | Q4_0 | llama.cpp (Vulkan) | 42.66 | 1075.94 | – | – | tg128 | 2025-11-07 | community | – | source |
| RTX 5060 Ti | llama 7B | Q4_0 | llama.cpp (Vulkan) | 42.47 | 2718.32 | – | – | tg128 | 2025-07-30 | community | – | source |
| M3 Ultra (60 GPU) | llama 7B | F16 | llama.cpp (Metal) | 42.24 | 1121.8 | – | – | – | 2023-11-22 | community | – | source |
| Arc A750 | llama 7B | Q4_0 | llama.cpp (Vulkan) | 42.15 | 312.51 | – | – | tg128 | 2025-08-18 | community | – | source |
| RTX 4050 | llama 7B | Q4_0 | llama.cpp (Vulkan) | 41.89 | 1154.28 | – | – | tg128 | 2025-01-13 | community | – | source |
| M2 Max (38 GPU) | llama 7B | Q8_0 | llama.cpp (Metal) | 41.83 | 677.91 | – | – | – | 2023-11-22 | community | – | source |
| Arc Pro B50 | llama 7B | Q4_0 | llama.cpp (Vulkan) | 41.82 | 581.78 | – | – | tg128 | 2026-05-30 | community | – | source |
| Arc Pro B50 | llama 7B | Q4_0 | llama.cpp (Vulkan) | 41.82 | 581.78 | – | – | tg128 | 2026-05-28 | community | – | source |
| Radeon RX 480 G | llama 7B | Q4_0 | llama.cpp (Vulkan) | 41.48 | 321.57 | – | – | tg128 | 2025-11-26 | community | – | source |
| Arc A770 | llama 7B | Q4_0 | llama.cpp (SYCL) | 41.48 | 697.13 | – | – | tg128 | 2026-08-31 | community | – | source |
| Arc A750 | llama 7B | Q4_0 | llama.cpp (Vulkan) | 41.43 | 665.38 | – | – | tg128 | 2025-08-18 | community | – | source |
| Arc A770 | llama 7B | Q4_0 | llama.cpp (Vulkan) | 41.19 | 257.53 | – | – | tg128 | 2025-07-30 | community | – | source |
| M2 Ultra (76 GPU) | llama 7B | F16 | llama.cpp (Metal) | 41.02 | 1401.85 | – | – | – | 2023-11-22 | community | – | source |
| Arc B50 | llama 7B | Q4_0 | llama.cpp (Vulkan) | 41.02 | 132.48 | – | – | tg128 | 2025-12-06 | community | – | source |
| A750 | llama 7B | Q4_0 | llama.cpp (OpenCL) | 40.7 | 234.63 | – | – | tg128 | 2025-11-07 | community | – | source |
| Arc A770 | llama 7B | Q4_0 | llama.cpp (Vulkan) | 40.62 | 1243.46 | – | – | tg128 | 2026-05-31 | community | – | source |
| Arc A770 | llama 7B | Q4_0 | llama.cpp (Vulkan) | 40.56 | 737.29 | – | – | tg128 | 2025-07-30 | community | – | source |
| Radeon RX 480 G | llama 7B | Q4_0 | llama.cpp (Vulkan) | 40.4 | 106.39 | – | – | tg128 | 2025-10-27 | community | – | source |
| Radeon RX 480 G | llama 7B | Q4_0 | llama.cpp (Vulkan) | 40.37 | 217.77 | – | – | tg128 | 2025-11-26 | community | – | source |
| M1 Max (32 GPU) | llama 7B | Q8_0 | llama.cpp (Metal) | 40.2 | 537.37 | – | – | – | 2023-11-22 | community | – | source |
| Radeon RX 570 S | llama 7B | Q4_0 | llama.cpp (Vulkan) | 40.18 | 229.23 | – | – | tg128 | 2026-04-25 | community | – | source |
| Arc Pro B50 | llama 7B | Q4_0 | llama.cpp (Vulkan) | 40.13 | 590.01 | – | – | tg128 | 2026-05-30 | community | – | source |
| Arc Pro B50 | llama 7B | Q4_0 | llama.cpp (Vulkan) | 40.13 | 590.01 | – | – | tg128 | 2026-05-28 | community | – | source |
| Radeon RX 470 G | llama 7B | Q4_0 | llama.cpp (Vulkan) | 40.11 | 207.16 | – | – | tg128 | 2026-07-09 | community | – | source |
| Arc B50 | llama 7B | Q4_0 | llama.cpp (Vulkan) | 39.99 | 193.5 | – | – | tg128 | 2025-12-06 | community | – | source |
| M2 Max (30 GPU) | llama 7B | Q8_0 | llama.cpp (Metal) | 39.97 | 540.15 | – | – | – | 2023-11-22 | community | – | source |
| Radeon RX 480 G | llama 7B | Q4_0 | llama.cpp (Vulkan) | 39.93 | 107.79 | – | – | tg128 | 2025-10-27 | community | – | source |
| M2 Ultra (60 GPU) | llama 7B | F16 | llama.cpp (Metal) | 39.86 | 1128.59 | – | – | – | 2023-11-22 | community | – | source |
| Radeon RX 470 G | llama 7B | Q4_0 | llama.cpp (Vulkan) | 39.85 | 206.61 | – | – | tg128 | 2026-05-27 | community | – | source |
| M3 Ultra (80 GPU) | llama 7B | F16 | llama.cpp (Metal) | 39.78 | 1538.34 | – | – | – | 2023-11-22 | community | – | source |
| Radeon RX 470 G | llama 7B | Q4_0 | llama.cpp (Vulkan) | 39.75 | 207.79 | – | – | tg128 | 2026-05-27 | community | – | source |
| Arc A770 | llama 7B | Q4_0 | llama.cpp (Vulkan) | 39.41 | 661.92 | – | – | tg128 | 2025-04-12 | community | – | source |
| Radeon RX 580 S | llama 7B | Q4_0 | llama.cpp (Vulkan) | 39.32 | 258.03 | – | – | tg128 | 2025-05-12 | community | – | source |
| Arc A770 | llama 7B | Q4_0 | llama.cpp (Vulkan) | 39.13 | 1209.43 | – | – | tg128 | 2026-05-31 | community | – | source |
| Arc A770 | llama 7B | Q4_0 | llama.cpp (SYCL) | 39.09 | 692.82 | – | – | tg128 | 2026-05-22 | community | – | source |
| Radeon RX 470 G | llama 7B | Q4_0 | llama.cpp (Vulkan) | 38.96 | 198.66 | – | – | tg128 | 2026-05-27 | community | – | source |
| M5 Pro | llama 7B | Q8_0 | llama.cpp (Metal) | 38.94 | 1302.59 | – | – | tg128 | 2026-09-24 | community | – | source |
| M5 Pro (20 GPU) | llama 7B | Q8_0 | llama.cpp (Metal) | 38.92 | 1553.86 | – | – | – | 2023-11-22 | community | – | source |
| M5 Pro | llama 7B | Q8_0 | llama.cpp (Metal) | 38.92 | 1553.86 | – | – | tg128 | 2026-08-31 | community | – | source |
| M2 Pro (19 GPU) | llama 7B | Q4_0 | llama.cpp (Metal) | 38.86 | 341.19 | – | – | – | 2023-11-22 | community | – | source |
| Arc A770 | llama 7B | Q4_0 | llama.cpp (Vulkan) | 38.8 | 255.59 | – | – | tg128 | 2025-07-30 | community | contradiction | source |
| Radeon RX 470 G | llama 7B | Q4_0 | llama.cpp (Vulkan) | 38.69 | 198.3 | – | – | tg128 | 2026-05-27 | community | – | source |
| Radeon RX 470 G | llama 7B | Q4_0 | llama.cpp (Vulkan) | 38.63 | 218.07 | – | – | tg128 | 2025-08-23 | community | – | source |
| GeForce RTX 3050 | llama 7B | Q4_0 | llama.cpp (CUDA) | 38.51 | 1258.25 | – | – | tg128 | 2026-07-15 | community | – | source |
| Radeon RX 9060 XT | llama 7B | Q4_0 | llama.cpp (Vulkan) | 38.37 | 1649.97 | – | – | tg128 | 2025-12-30 | community | – | source |
| Radeon RX 9060 XT | llama 7B | Q4_0 | llama.cpp (Vulkan) | 38.31 | 317.42 | – | – | tg128 | 2025-12-30 | community | – | source |
| Radeon RX 470 G | llama 7B | Q4_0 | llama.cpp (Vulkan) | 38.21 | 192.0 | – | – | tg128 | 2025-09-08 | community | – | source |
| Radeon RX 580 S | llama 7B | Q4_0 | llama.cpp (Vulkan) | 38.2 | 222.98 | – | – | tg128 | 2025-06-02 | community | – | source |
| A770 | llama 7B | Q4_0 | llama.cpp (SYCL) | 38.1 | 845.83 | – | – | tg128 | 2026-01-28 | community | contradiction | source |
| Arc A770 | llama 7B | Q4_0 | llama.cpp (Vulkan) | 38.07 | 706.67 | – | – | tg128 | 2025-07-30 | community | contradiction | source |
| Radeon RX 7600 XT | llama 7B | Q4_0 | llama.cpp (Vulkan) | 37.94 | 401.03 | – | – | tg128 | 2026-02-15 | community | contradiction | source |
| M2 Pro (16 GPU) | llama 7B | Q4_0 | llama.cpp (Metal) | 37.87 | 294.24 | – | – | – | 2023-11-22 | community | – | source |
| M1 Max (24 GPU) | llama 7B | Q8_0 | llama.cpp (Metal) | 37.81 | 405.87 | – | – | – | 2023-11-22 | community | – | source |
| M5 Pro | llama 7B | Q8_0 | llama.cpp (Metal) | 37.78 | 1304.66 | – | – | tg128 | 2026-04-14 | community | – | source |
| M5 Pro | llama 7B | Q8_0 | llama.cpp (Metal) | 37.7 | 964.18 | – | – | tg128 | 2026-04-14 | community | – | source |
| Radeon RX 6900 XT | llama 7B | Q4_0 | llama.cpp (Vulkan) | 37.59 | 23.61 | – | – | tg128 | 2025-10-26 | community | contradiction | source |
| Radeon RX 570 S | llama 7B | Q4_0 | llama.cpp (Vulkan) | 37.44 | 226.11 | – | – | tg128 | 2026-04-25 | community | – | source |
| A750 | llama 7B | Q4_0 | llama.cpp (OpenCL) | 37.42 | 162.56 | – | – | tg128 | 2025-11-07 | community | – | source |
| GeForce RTX 3050 | llama 7B | Q4_0 | llama.cpp (CUDA) | 37.39 | 1147.01 | – | – | tg128 | 2026-07-15 | community | – | source |
| Radeon RX 6900 XT | llama 7B | Q4_0 | llama.cpp (Vulkan) | 37.33 | 21.66 | – | – | tg128 | 2025-10-26 | community | contradiction | source |
| Radeon RX 480 G | llama 7B | Q4_0 | llama.cpp (Vulkan) | 37.29 | 214.61 | – | – | tg128 | 2025-11-26 | community | contradiction | source |
| Radeon RX 470 G | llama 7B | Q4_0 | llama.cpp (Vulkan) | 37.28 | 197.26 | – | – | tg128 | 2026-02-25 | community | – | source |
| Radeon RX 480 G | llama 7B | Q4_0 | llama.cpp (Vulkan) | 37.23 | 194.52 | – | – | tg128 | 2025-10-24 | community | – | source |
| Radeon RX 470 G | llama 7B | Q4_0 | llama.cpp (Vulkan) | 37.15 | 182.13 | – | – | tg128 | 2025-09-08 | community | – | source |
| M5 Max (40 GPU) | llama 7B | F16 | llama.cpp (Metal) | 37.11 | 3158.49 | – | – | – | 2023-11-22 | community | – | source |
| M5 Max | llama 7B | F16 | llama.cpp (Metal) | 37.11 | 3158.49 | – | – | tg128 | 2026-08-25 | community | – | source |
| M1 Ultra (64 GPU) | llama 7B | F16 | llama.cpp (Metal) | 37.01 | 1168.89 | – | – | – | 2023-11-22 | community | – | source |
| RTX 3090 | llama 7B | Q4_0 | llama.cpp (Vulkan) | 36.82 | 315.39 | – | – | tg128 | 2025-06-18 | community | – | source |
| Radeon RX 6900 XT | llama 7B | Q4_0 | llama.cpp (Vulkan) | 36.72 | 23.9 | – | – | tg128 | 2025-10-26 | community | contradiction | source |
| Arc A750 | llama 7B | Q4_0 | llama.cpp (SYCL) | 36.64 | 1616.11 | – | – | tg128 | 2025-01-08 | community | – | source |
| Radeon RX 480 G | llama 7B | Q4_0 | llama.cpp (Vulkan) | 36.49 | 201.82 | – | – | tg128 | 2025-10-24 | community | – | source |
| M1 Pro (16 GPU) | llama 7B | Q4_0 | llama.cpp (Metal) | 36.41 | 266.25 | – | – | – | 2023-11-22 | community | – | source |
| Arc A750 | llama 7B | Q4_0 | llama.cpp (SYCL) | 36.2 | 1631.21 | – | – | tg128 | 2025-11-07 | community | – | source |
| Quadro RTX 5000 | llama 7B | Q4_0 | llama.cpp (CUDA) | 36.11 | 549.43 | – | – | tg128 | 2026-05-25 | community | – | source |
| M2 Ultra | llama 7B | Q4_0 | llama.cpp (Vulkan) | 36.07 | 198.83 | – | – | tg128 | 2025-11-25 | community | – | source |
| Radeon RX 580 | llama 7B | Q4_0 | llama.cpp (Vulkan) | 35.79 | 180.49 | – | – | tg128 | 2025-11-26 | community | – | source |
| M6 | llama 7B 6.74B | Q4_0 | llama.cpp (Metal) | 35.74 | 988.93 | – | – | tg128 | 2026-10-01 | community | – | source |
| Radeon RX 580 | llama 7B | Q4_0 | llama.cpp (Vulkan) | 35.59 | 157.61 | – | – | tg128 | 2025-11-26 | community | – | source |
| M1 Pro (14 GPU) | llama 7B | Q4_0 | llama.cpp (Metal) | 35.52 | 232.55 | – | – | – | 2023-11-22 | community | – | source |
| M5 Max | llama 7B | F16 | llama.cpp (Metal) | 35.39 | 3189.9 | – | – | tg128 | 2026-06-22 | community | – | source |
| Radeon RX 470 G | llama 7B | Q4_0 | llama.cpp (Vulkan) | 35.14 | 197.94 | – | – | tg128 | 2025-08-23 | community | – | source |
| Radeon RX 470 G | llama 7B | Q4_0 | llama.cpp (Vulkan) | 34.89 | 185.73 | – | – | tg128 | 2025-08-22 | community | – | source |
| Radeon RX 9060 XT | llama 7B | Q4_0 | llama.cpp (Vulkan) | 34.72 | 389.52 | – | – | tg128 | 2025-06-13 | community | – | source |
| Radeon RX 480 G | llama 7B | Q4_0 | llama.cpp (Vulkan) | 34.71 | 248.66 | – | – | tg128 | 2025-09-21 | community | – | source |
| GeForce RTX 5060 Ti | llama 7B | Q4_0 | llama.cpp (Vulkan) | 34.71 | 189.5 | – | – | tg128 | 2026-05-08 | community | contradiction | source |
| Quadro RTX 5000 | llama 7B | Q4_0 | llama.cpp (CUDA) | 34.71 | 545.12 | – | – | tg128 | 2026-05-25 | community | – | source |
| T4 | llama 7B | Q4_0 | llama.cpp (Vulkan) | 34.66 | 734.28 | – | – | tg128 | 2025-07-12 | community | – | source |
| Radeon RX 470 G | llama 7B | Q4_0 | llama.cpp (Vulkan) | 34.65 | 179.01 | – | – | tg128 | 2025-08-22 | community | – | source |
| Arc A770 | llama 7B | Q4_0 | llama.cpp (SYCL) | 34.47 | 912.36 | – | – | tg128 | 2026-05-22 | community | – | source |
| Radeon RX 470 G | llama 7B | Q4_0 | llama.cpp (Vulkan) | 34.42 | 197.19 | – | – | tg128 | 2026-02-25 | community | – | source |
| Radeon RX 6600 | llama 7B | Q4_0 | llama.cpp (Vulkan) | 34.34 | 422.67 | – | – | tg128 | 2025-06-13 | community | – | source |
| M2 Ultra | llama 7B | Q4_0 | llama.cpp (Vulkan) | 34.34 | 205.98 | – | – | tg128 | 2025-11-25 | community | – | source |
| M3 Max (30 GPU) | llama 7B | Q8_0 | llama.cpp (Metal) | 34.3 | 566.4 | – | – | – | 2023-11-22 | community | – | source |
| GeForce RTX 5060 Ti | llama 7B | Q4_0 | llama.cpp (Vulkan) | 34.19 | 188.88 | – | – | tg128 | 2026-05-08 | community | contradiction | source |
| RTX 3090 | llama 7B | Q4_0 | llama.cpp (Vulkan) | 34.16 | 180.97 | – | – | tg128 | 2025-06-18 | community | – | source |
| M1 Ultra (48 GPU) | llama 7B | F16 | llama.cpp (Metal) | 33.92 | 875.81 | – | – | – | 2023-11-22 | community | – | source |
| RTX 3090 | llama 7B | Q4_0 | llama.cpp (Vulkan) | 33.9 | 186.24 | – | – | tg128 | 2025-06-18 | community | – | source |
| Radeon RX 470 G | llama 7B | Q4_0 | llama.cpp (Vulkan) | 33.45 | 161.47 | – | – | tg128 | 2025-08-22 | community | – | source |
| T4 | llama 7B | Q4_0 | llama.cpp (Vulkan) | 33.36 | 708.19 | – | – | tg128 | 2025-07-12 | community | – | source |
| Radeon RX 580 | llama 7B | Q4_0 | llama.cpp (Vulkan) | 33.19 | 182.18 | – | – | tg128 | 2025-11-26 | community | – | source |
| Radeon RX 9070 XT | llama 7B | Q4_0 | llama.cpp (ROCm) | 32.78 | 1728.29 | – | – | tg128 | 2026-04-01 | community | – | source |
| M5 (10 GPU) | llama 7B | Q4_0 | llama.cpp (Metal) | 31.88 | 722.79 | – | – | – | 2023-11-22 | community | – | source |
| M5 | llama 7B | Q4_0 | llama.cpp (Metal) | 31.88 | 722.79 | – | – | tg128 | 2026-08-30 | community | – | source |
| Arc B570 | llama 7B | Q4_0 | llama.cpp (Vulkan) | 31.74 | 285.34 | – | – | tg128 | 2025-08-17 | community | contradiction | source |
| M4 Max (40 GPU) | llama 7B | F16 | llama.cpp (Metal) | 31.64 | 922.83 | – | – | – | 2023-11-22 | community | – | source |
| M3 Pro | llama 7B | Q4_0 | llama.cpp (Metal) | 31.54 | 371.72 | – | – | tg128 | 2025-02-11 | community | – | source |
| Radeon RX 480 G | llama 7B | Q4_0 | llama.cpp (Vulkan) | 31.53 | 102.55 | – | – | tg128 | 2025-10-27 | community | contradiction | source |
| M3 Pro | llama 7B | Q4_0 | llama.cpp (Metal) | 31.52 | 368.92 | – | – | tg128 | 2025-02-11 | community | – | source |
| Arc A770 | llama 7B | Q4_0 | llama.cpp (SYCL) | 31.45 | 934.47 | – | – | tg128 | 2025-07-30 | community | – | source |
| Arc B570 | llama 7B | Q4_0 | llama.cpp (Vulkan) | 31.44 | 651.09 | – | – | tg128 | 2025-05-22 | community | – | source |
| Arc B570 | llama 7B | Q4_0 | llama.cpp (Vulkan) | 31.39 | 356.09 | – | – | tg128 | 2025-05-22 | community | – | source |
| Arc B570 | llama 7B | Q4_0 | llama.cpp (Vulkan) | 31.31 | 468.82 | – | – | tg128 | 2025-08-17 | community | contradiction | source |
| Radeon RX 480 G | llama 7B | Q4_0 | llama.cpp (Vulkan) | 31.23 | 103.24 | – | – | tg128 | 2025-10-27 | community | contradiction | source |
| M5 | llama 7B | Q4_0 | llama.cpp (Metal) | 31.02 | 636.36 | – | – | tg128 | 2025-10-30 | community | – | source |
| Arc B570 | llama 7B | Q4_0 | llama.cpp (Vulkan) | 30.89 | 226.82 | – | – | tg128 | 2025-05-22 | community | – | source |
| Arc B570 | llama 7B | Q4_0 | llama.cpp (Vulkan) | 30.88 | 342.71 | – | – | tg128 | 2025-05-22 | community | – | source |
| Radeon RX 470 G | llama 7B | Q4_0 | llama.cpp (Vulkan) | 30.77 | 147.84 | – | – | tg128 | 2025-01-09 | community | – | source |
| M3 Pro (18 GPU) | llama 7B | Q4_0 | llama.cpp (Metal) | 30.74 | 341.67 | – | – | – | 2023-11-22 | community | – | source |
| M4 Pro (20 GPU) | llama 7B | Q8_0 | llama.cpp (Metal) | 30.69 | 449.62 | – | – | – | 2023-11-22 | community | – | source |
| M5 | llama 7B | Q4_0 | llama.cpp (Metal) | 30.67 | 650.65 | – | – | tg128 | 2025-10-30 | community | – | source |
| M3 Pro (14 GPU) | llama 7B | Q4_0 | llama.cpp (Metal) | 30.65 | 269.49 | – | – | – | 2023-11-22 | community | – | source |
| M4 Pro (16 GPU) | llama 7B | Q8_0 | llama.cpp (Metal) | 30.54 | 367.13 | – | – | – | 2023-11-22 | community | – | source |
| NVIDIA RTX 3500 | llama 7B | Q4_0 | llama.cpp (CUDA) | 30.23 | 1406.43 | – | – | tg128 | 2025-09-20 | community | – | source |
| Radeon RX 470 G | llama 7B | Q4_0 | llama.cpp (Vulkan) | 29.52 | 155.22 | – | – | tg128 | 2025-08-22 | community | – | source |
| Ryzen AI Max+ 395 | llama 7B | Q4_0 | CPU | 29.52 | 340.11 | – | – | tg128 | 2026-10-02 | community | – | source |
| Radeon RX 470 G | llama 7B | Q4_0 | llama.cpp (Vulkan) | 29.28 | 140.52 | – | – | tg128 | 2025-08-22 | community | – | source |
| Radeon 8060S | llama 7B | Q4_0 | llama.cpp (CPU) | 29.28 | 304.53 | – | – | tg128 | 2025-11-12 | community | – | source |
| NVIDIA RTX 3500 | llama 7B | Q4_0 | llama.cpp (CUDA) | 28.75 | 1610.14 | – | – | tg128 | 2025-09-20 | community | – | source |
| Radeon RX 9060 XT | llama 7B | Q4_0 | llama.cpp (Vulkan) | 28.37 | 142.04 | – | – | tg128 | 2026-02-03 | community | contradiction | source |
| Radeon RX 6500 XT | llama 7B | Q4_0 | llama.cpp (Vulkan) | 27.81 | 255.25 | – | – | tg128 | 2025-05-04 | community | – | source |
| Ryzen AI Max+ 395 | llama 7B | Q4_0 | CPU | 27.68 | 300.27 | – | – | tg128 | 2026-10-02 | community | – | source |
| Radeon 8060S | llama 7B | Q4_0 | llama.cpp (CPU) | 27.58 | 308.17 | – | – | tg128 | 2025-11-12 | community | – | source |
| Arc A750 | llama 7B | Q4_0 | llama.cpp (Vulkan) | 27.57 | 88.86 | – | – | tg128 | 2025-01-08 | community | – | source |
| Radeon RX 580 S | llama 7B | Q4_0 | llama.cpp (Vulkan) | 27.51 | 212.32 | – | – | tg128 | 2025-06-02 | community | – | source |
| Radeon RX 480 G | llama 7B | Q4_0 | llama.cpp (Vulkan) | 27.42 | 237.27 | – | – | tg128 | 2025-09-21 | community | – | source |
| Arc A380 | llama 7B | Q4_0 | llama.cpp (SYCL) | 26.57 | 220.1 | – | – | tg128 | 2026-06-05 | community | – | source |
| M3 Pro | llama 7B | Q4_0 | llama.cpp (Vulkan) | 26.39 | 263.7 | – | – | tg128 | 2025-02-11 | community | – | source |
| M3 Pro | llama 7B | Q4_0 | llama.cpp (Vulkan) | 26.23 | 259.11 | – | – | tg128 | 2025-02-11 | community | – | source |
| M4 | llama 7B | F16 | llama.cpp (Metal) | 26.05 | 790.33 | – | – | tg128 | 2025-05-20 | community | – | source |
| Arc A380 | llama 7B | Q4_0 | llama.cpp (SYCL) | 25.81 | 213.39 | – | – | tg128 | 2026-06-05 | community | – | source |
| Arc A380 | llama 7B | Q4_0 | llama.cpp (SYCL) | 25.63 | 358.39 | – | – | tg128 | 2026-06-05 | community | – | source |
| Radeon RX 6600 | llama 7B | Q4_0 | llama.cpp (Vulkan) | 25.61 | 277.2 | – | – | tg128 | 2025-06-13 | community | – | source |
| Arc A380 | llama 7B | Q4_0 | llama.cpp (SYCL) | 25.57 | 362.69 | – | – | tg128 | 2026-06-05 | community | – | source |
| Arc(TM) 140V | llama 7B | Q4_0 | llama.cpp (Vulkan) | 25.12 | 867.27 | – | – | tg128 | 2026-09-13 | community | – | source |
| M3 Max (40 GPU) | llama 7B | F16 | llama.cpp (Metal) | 25.09 | 779.17 | – | – | – | 2023-11-22 | community | – | source |
| Radeon RX 6800M | llama 7B | Q4_0 | llama.cpp (Vulkan) | 24.76 | 435.39 | – | – | tg128 | 2025-07-17 | community | – | source |
| M2 Max (38 GPU) | llama 7B | F16 | llama.cpp (Metal) | 24.65 | 755.67 | – | – | – | 2023-11-22 | community | – | source |
| Radeon RX 570 S | llama 7B | Q4_0 | llama.cpp (Vulkan) | 24.63 | 53.68 | – | – | tg128 | 2026-04-25 | community | contradiction | source |
| Radeon RX 6800M | llama 7B | Q4_0 | llama.cpp (Vulkan) | 24.56 | 480.52 | – | – | tg128 | 2025-07-17 | community | – | source |
| Arc(TM) 140V | llama 7B | Q4_0 | llama.cpp (Vulkan) | 24.37 | 864.99 | – | – | tg128 | 2026-09-13 | community | – | source |
| M4 Max (32 GPU) | llama 7B | F16 | llama.cpp (Metal) | 24.29 | 736.25 | – | – | – | 2023-11-22 | community | – | source |
| M2 Max (30 GPU) | llama 7B | F16 | llama.cpp (Metal) | 24.16 | 600.46 | – | – | – | 2023-11-22 | community | – | source |
| M4 (10 GPU) | llama 7B | Q4_0 | llama.cpp (Metal) | 24.11 | 221.29 | – | – | – | 2023-11-22 | community | – | source |
| Radeon RX 570 S | llama 7B | Q4_0 | llama.cpp (Vulkan) | 24.11 | 52.66 | – | – | tg128 | 2026-04-25 | community | contradiction | source |
| GB10 | llama 7B | Q4_0 | llama.cpp (CPU) | 24.08 | 313.01 | – | – | tg128 | 2026-01-20 | community | – | source |
| M2 Max | llama 7B | Q4_0 | llama.cpp (Vulkan) | 24.07 | 79.94 | – | – | tg128 | 2024-06-21 | community | – | source |
| Arc A380 | llama 7B | Q4_0 | llama.cpp (SYCL) | 24.04 | 224.16 | – | – | tg128 | 2026-06-05 | community | – | source |
| Arc A380 | llama 7B | Q4_0 | llama.cpp (SYCL) | 23.97 | 491.34 | – | – | tg128 | 2026-06-05 | community | – | source |
| Arc A380 | llama 7B | Q4_0 | llama.cpp (SYCL) | 23.84 | 495.16 | – | – | tg128 | 2026-06-05 | community | – | source |
| Arc A380 | llama 7B | Q4_0 | llama.cpp (SYCL) | 23.81 | 226.11 | – | – | tg128 | 2026-06-05 | community | – | source |
| A4000 | llama 7B | Q4_0 | llama.cpp (CUDA) | 23.76 | 311.55 | – | – | tg128 | 2025-08-19 | community | – | source |
| A4000 | llama 7B | Q4_0 | llama.cpp (CUDA) | 23.63 | 309.3 | – | – | tg128 | 2025-08-19 | community | – | source |
| Arc A750 | llama 7B | Q4_0 | llama.cpp (SYCL) | 23.61 | 506.22 | – | – | tg128 | 2025-11-07 | community | – | source |
| M4 | llama 7B | Q4_0 | llama.cpp (Metal) | 23.29 | 212.13 | – | – | tg128 | 2024-05-20 | community | – | source |
| Radeon RX 5700 XT | llama 7B | Q4_0 | llama.cpp (Vulkan) | 23.21 | 214.48 | – | – | tg128 | 2025-01-12 | community | contradiction | source |
| Radeon 890M | llama 7B | Q4_0 | llama.cpp (Vulkan) | 23.1 | 373.78 | – | – | tg128 | 2025-12-14 | community | – | source |
| Radeon 890M | llama 7B | Q4_0 | llama.cpp (Vulkan) | 23.06 | 374.88 | – | – | tg128 | 2025-12-14 | community | – | source |
| M1 Max (32 GPU) | llama 7B | F16 | llama.cpp (Metal) | 23.03 | 599.53 | – | – | – | 2023-11-22 | community | – | source |
| M2 Pro (19 GPU) | llama 7B | Q8_0 | llama.cpp (Metal) | 23.01 | 344.5 | – | – | – | 2023-11-22 | community | – | source |
| M2 Max | llama 7B | Q4_0 | llama.cpp (CPU) | 22.86 | 49.04 | – | – | tg128 | 2024-06-21 | community | – | source |
| RTX 3090 | llama 7B | Q4_0 | llama.cpp (Vulkan) | 22.8 | 307.71 | – | – | tg128 | 2025-06-18 | community | – | source |
| M2 Pro (16 GPU) | llama 7B | Q8_0 | llama.cpp (Metal) | 22.7 | 288.46 | – | – | – | 2023-11-22 | community | – | source |
| M1 Max (24 GPU) | llama 7B | F16 | llama.cpp (Metal) | 22.55 | 453.03 | – | – | – | 2023-11-22 | community | – | source |
| RTX 3090 | llama 7B | Q4_0 | llama.cpp (Vulkan) | 22.41 | 311.56 | – | – | tg128 | 2025-06-18 | community | – | source |
| Radeon 890M | llama 7B | Q4_0 | llama.cpp (ROCm) | 22.37 | 545.1 | – | – | tg128 | 2025-12-14 | community | – | source |
| M1 Pro (16 GPU) | llama 7B | Q8_0 | llama.cpp (Metal) | 22.34 | 270.37 | – | – | – | 2023-11-22 | community | – | source |
| Radeon 890M | llama 7B | Q4_0 | llama.cpp (Vulkan) | 22.03 | 318.6 | – | – | tg128 | 2025-12-14 | community | – | source |
| Radeon 890M | llama 7B | Q4_0 | llama.cpp (Vulkan) | 21.99 | 329.76 | – | – | tg128 | 2025-12-14 | community | – | source |
| M1 Pro (14 GPU) | llama 7B | Q8_0 | llama.cpp (Metal) | 21.95 | 235.16 | – | – | – | 2023-11-22 | community | – | source |
| M2 (10 GPU) | llama 7B | Q4_0 | llama.cpp (Metal) | 21.91 | 179.57 | – | – | – | 2023-11-22 | community | – | source |
| M2 (8 GPU) | llama 7B | Q4_0 | llama.cpp (Metal) | 21.7 | 145.91 | – | – | – | 2023-11-22 | community | – | source |
| M5 Pro (20 GPU) | llama 7B | F16 | llama.cpp (Metal) | 21.55 | 1588.78 | – | – | – | 2023-11-22 | community | – | source |
| M5 Pro | llama 7B | F16 | llama.cpp (Metal) | 21.55 | 1588.78 | – | – | tg128 | 2026-08-31 | community | – | source |
| M3 | llama 7B | Q4_0 | llama.cpp (ROCm) | 21.47 | 310.95 | – | – | tg128 | 2024-01-21 | community | – | source |
| M5 Pro | llama 7B | F16 | llama.cpp (Metal) | 21.39 | 1310.78 | – | – | tg128 | 2026-09-24 | community | – | source |
| M5 Pro | llama 7B | F16 | llama.cpp (Metal) | 21.37 | 695.51 | – | – | tg128 | 2026-04-14 | community | – | source |
| M3 (10 GPU) | llama 7B | Q4_0 | llama.cpp (Metal) | 21.34 | 186.75 | – | – | – | 2023-11-22 | community | – | source |
| Radeon 780M | llama 7B | Q4_0 | llama.cpp (Vulkan) | 21.28 | 311.02 | – | – | tg128 | 2026-05-01 | community | – | source |
| Radeon 680M | llama 7B | Q4_0 | llama.cpp (Vulkan) | 21.26 | 240.89 | – | – | tg128 | 2025-10-19 | community | – | source |
| Radeon 890M | llama 7B | Q4_0 | llama.cpp (ROCm) | 21.18 | 424.24 | – | – | tg128 | 2025-12-14 | community | – | source |
| Radeon 680M | llama 7B | Q4_0 | llama.cpp (Vulkan) | 21.15 | 277.91 | – | – | tg128 | 2025-10-19 | community | – | source |
| M5 Pro | llama 7B | F16 | llama.cpp (Metal) | 21.11 | 1310.28 | – | – | tg128 | 2026-04-14 | community | – | source |
| M6 | llama 7B 6.74B | Q8_0 | llama.cpp (Metal) | 21.01 | 970.46 | – | – | tg128 | 2026-10-01 | community | – | source |
| M2 Pro | llama 7B | Q4_0 | llama.cpp (Vulkan) | 20.97 | 58.86 | – | – | tg128 | 2025-08-17 | community | – | source |
| M2 Pro | llama 7B | Q4_0 | llama.cpp (Vulkan) | 20.95 | 62.7 | 36.5 | – | tg128 | 2025-08-17 | community | – | source |
| Radeon 780M | llama 7B | Q4_0 | llama.cpp (Vulkan) | 20.59 | 297.39 | – | – | tg128 | 2025-11-04 | community | – | source |
| Radeon 780M | llama 7B | Q4_0 | llama.cpp (Vulkan) | 20.53 | 266.19 | – | – | tg128 | 2025-11-04 | community | – | source |
| Arc A380 | llama 7B | Q4_0 | llama.cpp (SYCL) | 20.26 | 227.06 | – | – | tg128 | 2026-06-13 | community | – | source |
| RTX 3060 | llama 7B | Q4_0 | llama.cpp (Vulkan) | 20.18 | 254.09 | – | – | tg128 | 2025-03-20 | community | – | source |
| Radeon 780M | llama 7B | Q4_0 | llama.cpp (Vulkan) | 20.1 | 245.79 | – | – | tg128 | 2025-02-09 | community | – | source |
| Radeon 890M | llama 7B | Q4_0 | llama.cpp (Vulkan) | 20.0 | 291.42 | – | – | tg128 | 2026-01-17 | community | – | source |
| Radeon 890M | llama 7B | Q4_0 | llama.cpp (Vulkan) | 19.93 | 257.01 | – | – | tg128 | 2025-08-14 | community | – | source |
| Radeon 780M | llama 7B | Q4_0 | llama.cpp (Vulkan) | 19.91 | 281.62 | – | – | tg128 | 2025-03-17 | community | – | source |
| Radeon 780M | llama 7B | Q4_0 | llama.cpp (Vulkan) | 19.81 | 267.27 | – | – | tg128 | 2026-05-01 | community | – | source |
| Radeon 780M | llama 7B | Q4_0 | llama.cpp (Vulkan) | 19.81 | 349.57 | – | – | tg128 | 2025-11-04 | community | – | source |
| Radeon 890M | llama 7B | Q4_0 | llama.cpp (Vulkan) | 19.81 | 264.67 | – | – | tg128 | 2025-08-14 | community | – | source |
| M2 Max | llama 7B | Q8_0 | llama.cpp (Vulkan) | 19.79 | 79.79 | – | – | tg128 | 2024-06-21 | community | – | source |
| Radeon 890M | llama 7B | Q4_0 | llama.cpp (Vulkan) | 19.78 | 314.97 | – | – | tg128 | 2026-03-05 | community | – | source |
| Arc A380 | llama 7B | Q4_0 | llama.cpp (SYCL) | 19.76 | 441.14 | – | – | tg128 | 2026-06-13 | community | – | source |
| Arc A380 | llama 7B | Q4_0 | llama.cpp (SYCL) | 19.7 | 216.06 | – | – | tg128 | 2026-06-05 | community | – | source |
| Radeon 780M | llama 7B | Q4_0 | llama.cpp (Vulkan) | 19.66 | 326.84 | – | – | tg128 | 2025-11-04 | community | – | source |
| M3 Max (30 GPU) | llama 7B | F16 | llama.cpp (Metal) | 19.54 | 589.41 | – | – | – | 2023-11-22 | community | – | source |
| Radeon 890M | llama 7B | Q4_0 | llama.cpp (Vulkan) | 19.39 | 354.86 | – | – | tg128 | 2025-09-21 | community | – | source |
| Radeon 890M | llama 7B | Q4_0 | llama.cpp (Vulkan) | 19.29 | 380.03 | – | – | tg128 | 2025-09-21 | community | – | source |
| Arc A380 | llama 7B | Q4_0 | llama.cpp (OpenVINO) | 19.21 | 1513.67 | – | – | tg128 | 2026-06-13 | community | – | source |
| Radeon 890M | llama 7B | Q4_0 | llama.cpp (Vulkan) | 19.16 | 278.49 | 34 | – | tg128 | 2026-01-17 | community | – | source |
| Arc A380 | llama 7B | Q4_0 | llama.cpp (SYCL) | 19.12 | 351.23 | – | – | tg128 | 2026-06-05 | community | – | source |
| Radeon 890M | llama 7B | Q4_0 | llama.cpp (ROCm) | 19.08 | 110.39 | – | – | tg128 | 2026-03-05 | community | – | source |
| Radeon 890M | llama 7B | Q4_0 | llama.cpp (ROCm) | 19.06 | 531.03 | – | – | tg128 | 2026-01-17 | community | – | source |
| Radeon 840M | llama 7B | Q4_0 | llama.cpp (Vulkan) | 19.03 | 150.12 | – | – | tg128 | 2026-08-07 | community | – | source |
| Radeon RX 580 | llama 7B | Q4_0 | llama.cpp (Vulkan) | 18.92 | 102.86 | – | – | tg128 | 2026-06-21 | community | – | source |
| Radeon 760M | llama 7B | Q4_0 | llama.cpp (Vulkan) | 18.89 | 219.24 | – | – | tg128 | 2026-04-25 | community | – | source |
| Radeon 890M | llama 7B | Q4_0 | llama.cpp (Vulkan) | 18.76 | 287.18 | – | – | tg128 | 2026-03-05 | community | – | source |
| Arc A380 | llama 7B | Q4_0 | llama.cpp (SYCL) | 18.62 | 351.23 | – | – | tg128 | 2026-06-13 | community | – | source |
| Arc A380 | llama 7B | Q4_0 | llama.cpp (SYCL) | 18.54 | 203.17 | – | – | tg128 | 2026-06-13 | community | – | source |
| M5 (10 GPU) | llama 7B | Q8_0 | llama.cpp (Metal) | 18.42 | 715.42 | – | – | – | 2023-11-22 | community | – | source |
| M5 | llama 7B | Q8_0 | llama.cpp (Metal) | 18.42 | 715.42 | – | – | tg128 | 2026-08-30 | community | – | source |
| AMD Custom GPU 0932 | llama 7B 6.74B | Q4_0 | llama.cpp (Vulkan) | 18.40 | 164.34 | – | – | tg128 | 2026-08-20 | community | – | source |
| RTX 3080 | llama 7B | Q4_0 | llama.cpp (Vulkan) | 18.32 | 230.79 | – | – | tg128 | 2026-04-03 | community | – | source |
| Radeon 890M | llama 7B | Q4_0 | llama.cpp (ROCm) | 18.24 | 121.84 | – | – | tg128 | 2026-03-05 | community | – | source |
| Radeon 780M | llama 7B | Q4_0 | llama.cpp (Vulkan) | 18.22 | 237.73 | – | – | tg128 | 2025-03-01 | community | – | source |
| Arc A770 | llama 7B | Q4_0 | llama.cpp (SYCL) | 18.22 | 288.17 | – | – | tg128 | 2025-07-30 | community | – | source |
| Arc A380 | llama 7B | Q4_0 | llama.cpp (SYCL) | 18.19 | 478.53 | – | – | tg128 | 2026-06-05 | community | – | source |
| Arc A380 | llama 7B | Q4_0 | llama.cpp (OpenVINO) | 18.18 | 1497.21 | – | – | tg128 | 2026-06-12 | community | – | source |
| Arc A380 | llama 7B | Q4_0 | llama.cpp (SYCL) | 18.16 | 220.98 | – | – | tg128 | 2026-06-05 | community | – | source |
| Radeon 890M | llama 7B | Q4_0 | llama.cpp (ROCm) | 18.02 | 437.1 | – | – | tg128 | 2026-01-17 | community | – | source |
| Arc A380 | llama 7B | Q4_0 | llama.cpp (Vulkan) | 17.87 | 263.7 | – | – | tg128 | 2026-06-13 | community | – | source |
| Quadro RTX 5000 | llama 7B | Q4_0 | llama.cpp (CUDA) | 17.84 | 218.32 | – | – | tg128 | 2026-05-25 | community | contradiction | source |
| Radeon 760M | llama 7B | Q4_0 | llama.cpp (Vulkan) | 17.69 | 187.02 | – | – | tg128 | 2026-04-25 | community | – | source |
| Radeon 840M | llama 7B | Q4_0 | llama.cpp (Vulkan) | 17.66 | 131.33 | – | – | tg128 | 2026-08-07 | community | – | source |
| Arc A380 | llama 7B | Q4_0 | llama.cpp (Vulkan) | 17.57 | 259.51 | – | – | tg128 | 2026-06-12 | community | – | source |
| M5 | llama 7B | Q8_0 | llama.cpp (Metal) | 17.55 | 512.15 | – | – | tg128 | 2025-10-30 | community | – | source |
| Arc A380 | llama 7B | Q4_0 | llama.cpp (Vulkan) | 17.55 | 249.81 | – | – | tg128 | 2026-06-13 | community | – | source |
| M3 Pro (18 GPU) | llama 7B | Q8_0 | llama.cpp (Metal) | 17.53 | 344.66 | – | – | – | 2023-11-22 | community | – | source |
| AMD Custom GPU 0932 | llama 7B 6.74B | Q4_0 | llama.cpp (Vulkan) | 17.52 | 144.31 | – | – | tg128 | 2026-08-20 | community | – | source |
| M5 | llama 7B | Q8_0 | llama.cpp (Metal) | 17.5 | 489.78 | – | – | tg128 | 2025-10-30 | community | – | source |
| M3 Pro (14 GPU) | llama 7B | Q8_0 | llama.cpp (Metal) | 17.44 | 272.11 | – | – | – | 2023-11-22 | community | – | source |
| Arc A380 | llama 7B | Q4_0 | llama.cpp (Vulkan) | 17.25 | 247.03 | – | – | tg128 | 2026-06-12 | community | – | source |
| M4 Pro (16 GPU) | llama 7B | F16 | llama.cpp (Metal) | 17.19 | 381.14 | – | – | – | 2023-11-22 | community | – | source |
| M4 Pro (20 GPU) | llama 7B | F16 | llama.cpp (Metal) | 17.18 | 464.48 | – | – | – | 2023-11-22 | community | – | source |
| Radeon RX 580 | llama 7B | Q4_0 | llama.cpp (Vulkan) | 17.0 | 96.83 | – | – | tg128 | 2026-06-21 | community | – | source |
| Quadro RTX 5000 | llama 7B | Q4_0 | llama.cpp (CUDA) | 16.74 | 216.34 | – | – | tg128 | 2026-05-25 | community | contradiction | source |
| Radeon 890M | llama 7B | Q4_0 | llama.cpp (ROCm) | 16.73 | 421.3 | – | – | tg128 | 2026-05-13 | community | – | source |
| Radeon 660M | llama 7B | Q4_0 | llama.cpp (Vulkan) | 16.64 | 138.61 | – | – | tg128 | 2025-10-19 | community | – | source |
| Radeon 780M | llama 7B | Q4_0 | llama.cpp (Vulkan) | 16.6 | 205.15 | – | – | tg128 | 2025-03-17 | community | contradiction | source |
| A770 | llama 7B | Q4_0 | llama.cpp (SYCL) | 16.57 | 437.75 | – | – | tg128 | 2026-01-28 | community | contradiction | source |
| RTX 3090 | llama 7B | Q4_0 | llama.cpp (Vulkan) | 16.56 | 75.59 | – | – | tg128 | 2025-06-18 | community | contradiction | source |
| Radeon 660M | llama 7B | Q4_0 | llama.cpp (Vulkan) | 16.56 | 148.05 | – | – | tg128 | 2025-10-19 | community | – | source |
| Radeon 660M | llama 7B | Q4_0 | llama.cpp (Vulkan) | 16.2 | 136.56 | – | – | tg128 | 2025-09-09 | community | – | source |
| Arc A380 | llama 7B | Q4_0 | llama.cpp (SYCL) | 16.17 | 276.43 | – | – | tg128 | 2026-06-12 | community | – | source |
| Radeon 890M | llama 7B | Q4_0 | llama.cpp (ROCm) | 16.16 | 213.31 | – | – | tg128 | 2025-10-28 | community | – | source |
| Radeon 840M | llama 7B | Q4_0 | llama.cpp (ROCm) | 16.12 | 172.58 | – | – | tg128 | 2026-08-07 | community | – | source |
| Radeon 890M | llama 7B | Q4_0 | llama.cpp (ROCm) | 16.12 | 417.37 | – | – | tg128 | 2026-05-13 | community | – | source |
| Radeon 660M | llama 7B | Q4_0 | llama.cpp (Vulkan) | 16.11 | 147.9 | – | – | tg128 | 2025-09-09 | community | – | source |
| Radeon RX 580 | llama 7B | Q4_0 | llama.cpp (Vulkan) | 16.08 | 100.35 | – | – | tg128 | 2025-08-27 | community | – | source |
| Radeon RX 7600 XT | llama 7B | Q4_0 | llama.cpp (Vulkan) | 16.06 | 49.49 | – | – | tg128 | 2026-02-15 | community | contradiction outlier | source |
| RTX 3090 | llama 7B | Q4_0 | llama.cpp (Vulkan) | 16.05 | 78.79 | – | – | tg128 | 2025-06-18 | community | contradiction | source |
| Arc A380 | llama 7B | Q4_0 | llama.cpp (SYCL) | 15.97 | 170.14 | – | – | tg128 | 2026-06-12 | community | – | source |
| Radeon 680M | llama 7B | Q4_0 | llama.cpp (Vulkan) | 15.42 | 325.54 | – | – | tg128 | 2026-10-10 | community | – | source |
| Radeon 890M | llama 7B | Q4_0 | llama.cpp (Vulkan) | 15.11 | 246.65 | – | – | tg128 | 2025-10-27 | community | – | source |
| Radeon RX 9060 XT | llama 7B | Q4_0 | llama.cpp (Vulkan) | 15.07 | 103.01 | – | – | tg128 | 2026-02-03 | community | contradiction | source |
| Arc A380 | llama 7B | Q4_0 | llama.cpp (SYCL) | 14.96 | 201.62 | – | – | tg128 | 2026-06-12 | community | – | source |
| Arc A380 | llama 7B | Q4_0 | llama.cpp (SYCL) | 14.95 | 347.28 | – | – | tg128 | 2026-06-12 | community | – | source |
| Radeon 890M | llama 7B | Q4_0 | llama.cpp (Vulkan) | 14.94 | 201.59 | – | – | tg128 | 2025-10-27 | community | – | source |
| Radeon 680M | llama 7B | Q4_0 | llama.cpp (Vulkan) | 14.7 | 1494.78 | – | – | tg128 | 2026-02-19 | community | – | source |
| Radeon 680M | llama 7B | Q4_0 | llama.cpp (Vulkan) | 14.58 | 260.18 | – | – | tg128 | 2026-10-10 | community | – | source |
| Radeon 840M | llama 7B | Q4_0 | llama.cpp (ROCm) | 14.33 | 134.09 | – | – | tg128 | 2026-08-07 | community | – | source |
| M1 (7 GPU) | llama 7B | Q4_0 | llama.cpp (Metal) | 14.19 | 107.81 | – | – | – | 2023-11-22 | community | – | source |
| M1 (8 GPU) | llama 7B | Q4_0 | llama.cpp (Metal) | 14.15 | 117.96 | – | – | – | 2023-11-22 | community | – | source |
| M1 | llama 7B | Q4_0 | llama.cpp (Metal) | 13.99 | 131.46 | – | – | tg128 | 2025-05-20 | community | – | source |
| Radeon RX 580 | llama 7B | Q4_0 | llama.cpp (Vulkan) | 13.99 | 85.37 | – | – | tg128 | 2025-08-27 | community | – | source |
| Radeon 680M | llama 7B | Q4_0 | llama.cpp (Vulkan) | 13.86 | 670.41 | – | – | tg128 | 2026-02-19 | community | – | source |
| M2 Max | llama 7B | Q8_0 | llama.cpp (CPU) | 13.79 | 52.83 | – | – | tg128 | 2024-06-21 | community | – | source |
| M4 (10 GPU) | llama 7B | Q8_0 | llama.cpp (Metal) | 13.54 | 223.64 | – | – | – | 2023-11-22 | community | – | source |
| M2 | llama 7B | Q4_0 | llama.cpp (Vulkan) | 13.5 | 50.79 | – | – | tg128 | 2025-11-07 | community | – | source |
| A770 | llama 7B | Q4_0 | llama.cpp (SYCL) | 13.23 | 480.14 | – | – | tg128 | 2025-09-21 | community | contradiction | source |
| Radeon 840M | llama 7B | Q4_0 | llama.cpp (CPU) | 13.18 | 61.42 | – | – | tg128 | 2026-08-07 | community | – | source |
| M2 Pro (19 GPU) | llama 7B | F16 | llama.cpp (Metal) | 13.06 | 384.38 | – | – | – | 2023-11-22 | community | – | source |
| Radeon 890M | llama 7B | Q4_0 | llama.cpp (ROCm) | 12.97 | 163.25 | – | – | tg128 | 2025-10-28 | community | – | source |
| Radeon 840M | llama 7B | Q4_0 | llama.cpp (CPU) | 12.95 | 60.33 | – | – | tg128 | 2026-08-07 | community | – | source |
| M1 | llama 7B | Q4_0 | llama.cpp (Vulkan) | 12.85 | 35.93 | – | – | tg128 | 2025-11-21 | community | – | source |
| M1 Pro (14 GPU) | llama 7B | F16 | llama.cpp (Metal) | 12.75 | 262.65 | – | – | – | 2023-11-22 | community | – | source |
| M1 Pro (16 GPU) | llama 7B | F16 | llama.cpp (Metal) | 12.75 | 302.14 | – | – | – | 2023-11-22 | community | – | source |
| Radeon RX 580 S | llama 7B | Q4_0 | llama.cpp (Vulkan) | 12.72 | 131.42 | – | – | tg128 | 2025-09-07 | community | – | source |
| Radeon 660M | llama 7B | Q4_0 | llama.cpp (ROCm) | 12.72 | 120.52 | – | – | tg128 | 2025-12-10 | community | – | source |
| M2 Pro (16 GPU) | llama 7B | F16 | llama.cpp (Metal) | 12.47 | 312.65 | – | – | – | 2023-11-22 | community | – | source |
| M1 | llama 7B | Q4_0 | llama.cpp (Vulkan) | 12.47 | 38.29 | – | – | tg128 | 2025-11-21 | community | – | source |
| M1 | llama 7B | Q4_0 | llama.cpp (Vulkan) | 12.41 | 31.31 | 20.35 | – | tg128 | 2025-08-17 | community | – | source |
| M1 | llama 7B | Q4_0 | llama.cpp (Vulkan) | 12.38 | 28.65 | – | – | tg128 | 2025-08-17 | community | – | source |
| M3 (10 GPU) | llama 7B | Q8_0 | llama.cpp (Metal) | 12.27 | 187.52 | – | – | – | 2023-11-22 | community | – | source |
| M2 | llama 7B | Q4_0 | llama.cpp (Vulkan) | 12.25 | 46.81 | – | – | tg128 | 2025-11-07 | community | – | source |
| M2 (10 GPU) | llama 7B | Q8_0 | llama.cpp (Metal) | 12.21 | 181.4 | – | – | – | 2023-11-22 | community | – | source |
| M2 (8 GPU) | llama 7B | Q8_0 | llama.cpp (Metal) | 12.18 | 147.27 | – | – | – | 2023-11-22 | community | – | source |
| Radeon RX 6600 | llama 7B | Q4_0 | llama.cpp (Vulkan) | 12.16 | 26.05 | – | – | tg128 | 2025-03-10 | community | – | source |
| Radeon RX 6600 | llama 7B | Q4_0 | llama.cpp (CPU) | 12.08 | 26.26 | – | – | tg128 | 2025-03-10 | community | – | source |
| RTX 3060 | llama 7B | Q4_K - Medium | llama.cpp (Vulkan) | 11.8 | 218.01 | – | – | tg128 | 2025-11-18 | community | – | source |
| Radeon RX 580 S | llama 7B | Q4_0 | llama.cpp (Vulkan) | 11.62 | 177.33 | – | – | tg128 | 2025-09-07 | community | – | source |
| M6 | llama 7B 6.74B | F16 | llama.cpp (Metal) | 11.59 | 952.53 | – | – | tg128 | 2026-10-01 | community | – | source |
| A770 | llama 7B | Q4_0 | llama.cpp (SYCL) | 11.14 | 497.1 | – | – | tg128 | 2026-01-28 | community | contradiction | source |
| M2 | llama 7B | Q4_0 | llama.cpp (Vulkan) | 11.07 | 38.67 | – | – | tg128 | 2025-01-08 | community | – | source |
| RTX 3060 | llama 7B | Q4_0 | llama.cpp (Vulkan) | 10.81 | 90.15 | – | – | tg128 | 2025-03-20 | community | contradiction | source |
| Radeon RX 550 | llama 7B | Q4_0 | llama.cpp (Vulkan) | 10.2 | 52.66 | – | – | tg128 | 2025-11-16 | community | – | source |
| RTX 3090 | llama 7B | Q4_0 | llama.cpp (Vulkan) | 10.09 | 76.53 | – | – | tg128 | 2025-06-18 | community | contradiction | source |
| RTX 3090 | llama 7B | Q4_0 | llama.cpp (Vulkan) | 10.06 | 81.59 | – | – | tg128 | 2025-06-18 | community | contradiction outlier | source |
| M3 Pro (18 GPU) | llama 7B | F16 | llama.cpp (Metal) | 9.89 | 357.45 | – | – | – | 2023-11-22 | community | – | source |
| Radeon RX 6800 | llama 7B | Q4_0 | llama.cpp (Vulkan) | 9.87 | 113.32 | – | – | tg128 | 2026-02-12 | community | contradiction outlier | source |
| RTX 3060 | llama 7B | Q4_0 | llama.cpp (Vulkan) | 9.79 | 72.76 | – | – | tg128 | 2025-12-03 | community | – | source |
| Radeon RX 7600 | llama 7B | Q4_0 | llama.cpp (CPU) | 9.72 | 44.12 | – | – | tg128 | 2026-07-14 | community | – | source |
| Radeon RX 7600 | llama 7B | Q4_0 | llama.cpp (CPU) | 9.72 | 42.94 | – | – | tg128 | 2026-07-14 | community | – | source |
| M5 | llama 7B | F16 | llama.cpp (Metal) | 9.67 | 374.35 | – | – | tg128 | 2025-10-30 | community | – | source |
| Radeon RX 6800 | llama 7B | Q4_0 | llama.cpp (Vulkan) | 9.63 | 103.87 | – | – | tg128 | 2026-02-12 | community | contradiction outlier | source |
| M5 | llama 7B | F16 | llama.cpp (Metal) | 9.56 | 378.36 | – | – | tg128 | 2025-10-30 | community | – | source |
| RTX 3060 | llama 7B | Q4_0 | llama.cpp (Vulkan) | 9.48 | 74.29 | – | – | tg128 | 2025-12-03 | community | – | source |
| RTX 3060 | llama 7B | Q4_K - Medium | llama.cpp (Vulkan) | 9.4 | 1081.37 | – | – | tg128 | 2025-11-18 | community | – | source |
| RTX 3090 | llama 7B | Q4_0 | llama.cpp (Vulkan) | 8.55 | 117.48 | – | – | tg128 | 2025-06-18 | community | contradiction | source |
| Radeon 890M | llama 7B | Q4_0 | llama.cpp (Vulkan) | 8.31 | 293.41 | – | – | tg128 | 2025-10-27 | community | – | source |
| Radeon 890M | llama 7B | Q4_0 | llama.cpp (Vulkan) | 8.3 | 315.24 | – | – | tg128 | 2025-10-27 | community | contradiction | source |
| M2 Max | llama 7B | F16 | llama.cpp (CPU) | 8.19 | 26.18 | – | – | tg128 | 2024-06-21 | community | – | source |
| Radeon RX 7600 XT | llama 7B | Q4_0 | llama.cpp (Vulkan) | 8.17 | 34.95 | – | – | tg128 | 2026-02-15 | community | contradiction outlier | source |
| M1 (7 GPU) | llama 7B | Q8_0 | llama.cpp (Metal) | 7.92 | 108.21 | – | – | – | 2023-11-22 | community | – | source |
| M1 (8 GPU) | llama 7B | Q8_0 | llama.cpp (Metal) | 7.91 | 117.25 | – | – | – | 2023-11-22 | community | – | source |
| RTX 3060 | llama 7B | Q4_0 | llama.cpp (Vulkan) | 7.69 | 91.04 | – | – | tg128 | 2025-12-03 | community | – | source |
| M1 | llama 7B | Q8_0 | llama.cpp (Metal) | 7.67 | 133.34 | – | – | tg128 | 2025-05-20 | community | – | source |
| M4 (10 GPU) | llama 7B | F16 | llama.cpp (Metal) | 7.43 | 230.18 | – | – | – | 2023-11-22 | community | – | source |
| RTX 3060 | llama 7B | Q4_0 | llama.cpp (Vulkan) | 7.33 | 90.13 | – | – | tg128 | 2025-12-03 | community | – | source |
| M2 (10 GPU) | llama 7B | F16 | llama.cpp (Metal) | 6.72 | 201.34 | – | – | – | 2023-11-22 | community | – | source |
| RTX 3090 | llama 7B | Q4_0 | llama.cpp (Vulkan) | 5.82 | 16.62 | – | – | tg128 | 2025-06-18 | community | – | source |
| RTX 3090 | llama 7B | Q4_0 | llama.cpp (Vulkan) | 5.71 | 17.7 | – | – | tg128 | 2025-06-18 | community | – | source |
| RTX 3090 | llama 7B | Q4_0 | llama.cpp (Vulkan) | 5.61 | 42.12 | – | – | tg128 | 2025-06-18 | community | – | source |
| RTX 3090 | llama 7B | Q4_0 | llama.cpp (Vulkan) | 5.54 | 41.26 | – | – | tg128 | 2025-06-18 | community | – | source |
| RTX 3090 | llama 7B | Q4_0 | llama.cpp (Vulkan) | 5.42 | 15.22 | – | – | tg128 | 2025-06-18 | community | – | source |
| RTX 3090 | llama 7B | Q4_0 | llama.cpp (Vulkan) | 5.35 | 19.18 | – | – | tg128 | 2025-06-18 | community | – | source |
| RTX 3090 | llama 7B | Q4_0 | llama.cpp (Vulkan) | 2.87 | 84.19 | – | – | tg128 | 2025-06-18 | community | contradiction | source |
| RTX 3090 | llama 7B | Q4_0 | llama.cpp (Vulkan) | 2.85 | 107.8 | – | – | tg128 | 2025-06-18 | community | contradiction | source |
| RTX 3090 | llama 7B | Q4_0 | llama.cpp (Vulkan) | 2.67 | 23.19 | – | – | tg128 | 2025-06-18 | community | – | source |
| RTX 3090 | llama 7B | Q4_0 | llama.cpp (Vulkan) | 2.63 | 23.19 | – | – | tg128 | 2025-06-18 | community | – | source |
| M2 Max | llama 7B | F16 | llama.cpp (Vulkan) | 2.02 | 80.4 | – | – | tg128 | 2024-06-21 | community | – | source |