Llama 2 7B
18 records across 7 hardware.
| hardware | model | quant | backend | tok/s | pp tok/s | W | ttft s | ctx/tg | date | class | flags | source |
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| M2 Ultra (76 GPU) | Llama 2 7B | Q4_0 | llama.cpp (Metal) | 125.21 | 1488.81 | – | – | – | 2023-11-22 | community | – | source |
| M2 Ultra (76 GPU) | Llama 2 7B | Q4_0 | llama.cpp (Metal) | 109.41 | 1412.42 | – | – | – | 2023-11-22 | community | – | source |
| M2 Ultra (76 GPU) | Llama 2 7B | Q4_0 | llama.cpp (Metal) | 108.8 | 1391.78 | – | – | – | 2023-11-22 | community | – | source |
| M2 Ultra (76 GPU) | Llama 2 7B | Q4_0 | llama.cpp (Metal) | 94.27 | 1238.48 | – | – | – | 2023-11-22 | community | – | source |
| M2 Ultra (76-core GPU, 192 GB) | Llama 2 7B | Q4_0 | llama.cpp | 94.27 | 1238.48 | – | – | 512 | – | sourced | – | source |
| M2 Ultra (76 GPU) | Llama 2 7B | Q8_0 | llama.cpp (Metal) | 82.83 | 1486.51 | – | – | – | 2023-11-22 | community | – | source |
| 3090 | Llama 2 7B 7B | EXL2 4bpw | ExLlamaV2 | 80 | – | – | – | – | 2023-10 | community | – | source |
| M2 Ultra (76 GPU) | Llama 2 7B | Q8_0 | llama.cpp (Metal) | 73.35 | 1386.97 | – | – | – | 2023-11-22 | community | – | source |
| M2 Ultra (76 GPU) | Llama 2 7B | Q8_0 | llama.cpp (Metal) | 73.11 | 1368.18 | – | – | – | 2023-11-22 | community | – | source |
| M2 Ultra (76 GPU) | Llama 2 7B | Q8_0 | llama.cpp (Metal) | 66.64 | 1248.59 | – | – | – | 2023-11-22 | community | – | source |
| M3 Max (40-core GPU, 48 GB) | Llama 2 7B | Q4_0 | llama.cpp | 65.85 | 690.99 | – | – | 512 | – | sourced | – | source |
| M2 Ultra (76 GPU) | Llama 2 7B | F16 | llama.cpp (Metal) | 49.55 | 1627.82 | – | – | – | 2023-11-22 | community | – | source |
| M2 Ultra (76 GPU) | Llama 2 7B | F16 | llama.cpp (Metal) | 43.24 | 1561.35 | – | – | – | 2023-11-22 | community | – | source |
| M2 Ultra (76 GPU) | Llama 2 7B | F16 | llama.cpp (Metal) | 43.15 | 1525.95 | – | – | – | 2023-11-22 | community | – | source |
| M2 Ultra (76 GPU) | Llama 2 7B | F16 | llama.cpp (Metal) | 41.02 | 1401.85 | – | – | – | 2023-11-22 | community | – | source |
| M1 Pro (16-core GPU) | Llama 2 7B | Q4_0 | llama.cpp | 36.41 | 266.25 | – | – | 512 | – | sourced | – | source |
| M3 Pro (18-core GPU) | Llama 2 7B | Q4_0 | llama.cpp | 30.74 | 341.67 | – | – | 512 | – | sourced | – | source |
| M4 (10-core GPU, 16 GB) | Llama 2 7B | Q4_0 | llama.cpp | 24.11 | 221.29 | – | – | 512 | – | sourced | – | source |