Gemma-4-31B
81 records across 18 hardware.
| hardware | model | quant | backend | tok/s | pp tok/s | W | ttft s | ctx/tg | date | class | flags | source |
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| H200 | Gemma-4-31B 31B dense | as-published | vLLM | 1364.4 | – | – | – | – | 2026-07 | community | – | source |
| H200 | Gemma-4-31B 31B dense | as-published | vLLM | 945.3 | – | – | – | – | 2026-07 | community | – | source |
| H200 | Gemma-4-31B 31B dense | as-published | vLLM | 525.4 | – | – | – | – | 2026-07 | community | – | source |
| H200 | Gemma-4-31B 31B dense | as-published | vLLM | 280.4 | – | – | – | – | 2026-07 | community | – | source |
| H200 | Gemma-4-31B 31B dense | as-published | vLLM | 73.9 | – | – | – | – | 2026-07 | community | – | source |
| H200 | Gemma-4-31B 31B dense | as-published | vLLM | 44.3 | – | – | – | – | 2026-07 | community | – | source |
| M5 Ultra | Gemma-4-31B | 4bit | oMLX | 42.3 | 899.5 | – | – | 1024 | 2026-10-16 | community | – | source |
| M5 Ultra | Gemma-4-31B | 4bit | oMLX | 39.3 | 958.4 | – | – | 4096 | 2026-10-16 | community | – | source |
| M4 Max | Gemma-4-31B | 4bit | oMLX | 29.7 | 191.3 | – | – | 8192 | 2026-10-16 | community | – | source |
| M5 Max (128 GB) | Gemma-4-31B | Q4_K - Medium | MLX | 26 | – | – | 0.7 | – | – | sourced | – | source |
| M5 Max (64 GB) | Gemma-4-31B | Q4_K - Medium | Ollama | 26 | – | – | 0.9 | – | – | sourced | – | source |
| M5 Max | Gemma-4-31B | 4bit | oMLX | 25.0 | 644.0 | – | – | 4096 | 2026-10-13 | community | – | source |
| M5 Max | Gemma-4-31B | 4bit | oMLX | 24.8 | 705.8 | – | – | 1024 | 2026-10-13 | community | – | source |
| M2 Ultra | Gemma-4-31B | 4bit | oMLX | 24.8 | 174.5 | – | – | 4096 | 2026-10-15 | community | – | source |
| gfx1151 | Gemma-4-31B | AWQ-4bit | vLLM | 24.2 | – | – | – | 8192 | 2026-10-13 | sourced | – | source |
| M4 Max (48 GB) | Gemma-4-31B | Q4_K - Medium | MLX | 24 | – | – | 1.1 | – | – | sourced | – | source |
| M4 Max | Gemma-4-31B | 4bit | oMLX | 22.9 | 166.9 | – | – | 32768 | 2026-10-16 | community | – | source |
| gfx1151 | Gemma-4-31B | AWQ-4bit | vLLM | 22.5 | – | – | – | 8192 | 2026-10-13 | sourced | – | source |
| M5 Max | Gemma-4-31B | 5bit | oMLX | 21.8 | 603.0 | – | – | 1024 | 2026-10-15 | community | – | source |
| M2 Ultra | Gemma-4-31B | 4bit | oMLX | 21.6 | 171.6 | – | – | 16384 | 2026-10-15 | community | – | source |
| M5 Max | Gemma-4-31B | 5bit | oMLX | 21.3 | 636.6 | – | – | 4096 | 2026-10-15 | community | – | source |
| M3 Max | Gemma-4-31B | 4bit | oMLX | 18.0 | 179.9 | – | – | 1024 | 2026-10-12 | community | – | source |
| M3 Max | Gemma-4-31B | 4bit | oMLX | 17.9 | 175.9 | – | – | 4096 | 2026-10-12 | community | – | source |
| M4 Max | Gemma-4-31B | 4bit | oMLX | 17.7 | 162.4 | – | – | 4096 | 2026-10-11 | community | – | source |
| M4 Max | Gemma-4-31B | 4bit | oMLX | 17.7 | 162.4 | – | – | 4096 | 2026-10-11 | community | – | source |
| M5 Ultra | Gemma-4-31B | 4bit | oMLX | 17.4 | 799.4 | – | – | 65536 | 2026-10-16 | community | – | source |
| M4 Max | Gemma-4-31B | 4bit | oMLX | 16.7 | 153.5 | – | – | 16384 | 2026-10-11 | community | – | source |
| M4 Max | Gemma-4-31B | 4bit | oMLX | 16.7 | 153.5 | – | – | 16384 | 2026-10-11 | community | – | source |
| M5 Max | Gemma-4-31B | 8bit | oMLX | 15.5 | 643.6 | – | – | 1024 | 2026-10-07 | community | – | source |
| M5 Max | Gemma-4-31B | 8bit | oMLX | 15.2 | 639.7 | – | – | 1024 | 2026-10-16 | community | – | source |
| M5 Max | Gemma-4-31B | 8bit | oMLX | 15.1 | 691.6 | – | – | 4096 | 2026-10-07 | community | – | source |
| M4 Max | Gemma-4-31B | 5bit | oMLX | 14.8 | 159.0 | – | – | 1024 | 2026-10-07 | community | – | source |
| M5 Max | Gemma-4-31B | 8bit | oMLX | 14.5 | 586.5 | – | – | 1024 | 2026-10-16 | community | – | source |
| M4 Max | Gemma-4-31B | 5bit | oMLX | 14.3 | 147.3 | – | – | 4096 | 2026-10-07 | community | – | source |
| M5 Pro | Gemma-4-31B | 4bit | oMLX | 14.3 | 380.5 | – | – | 1024 | 2026-10-18 | community | – | source |
| M5 Pro | Gemma-4-31B | 4bit | oMLX | 14.1 | 311.0 | – | – | 4096 | 2026-10-10 | community | – | source |
| M5 Max | Gemma-4-31B | 8bit | oMLX | 13.5 | 626.1 | – | – | 4096 | 2026-10-16 | community | contradiction | source |
| M5 Pro | Gemma-4-31B | 4bit | oMLX | 13.4 | 389.3 | – | – | 4096 | 2026-10-18 | community | – | source |
| M5 Max | Gemma-4-31B | 8bit | oMLX | 13.3 | 635.0 | – | – | 8192 | 2026-10-16 | community | – | source |
| M5 Max | Gemma-4-31B | 5bit | oMLX | 12.7 | 384.0 | – | – | 65536 | 2026-10-15 | community | – | source |
| M5 Pro | Gemma-4-31B | 4bit | oMLX | 12.5 | 292.3 | – | – | 16384 | 2026-10-10 | community | – | source |
| M5 Max | Gemma-4-31B | 8bit | oMLX | 12.4 | 429.4 | – | – | 65536 | 2026-10-16 | community | – | source |
| M5 Max | Gemma-4-31B | 8bit | oMLX | 12.0 | 574.7 | – | – | 16384 | 2026-10-16 | community | – | source |
| M4 Pro (24 GB) | Gemma-4-31B | Q4_K - Medium | Ollama | 12 | – | – | 1.4 | – | – | sourced | – | source |
| M4 Pro | Gemma-4-31B | 4bit | oMLX | 11.9 | 103.8 | – | – | 1024 | 2026-10-12 | community | – | source |
| M1 Max | Gemma-4-31B | 4bit | oMLX | 11.4 | 66.6 | – | – | 1024 | 2026-10-06 | community | – | source |
| M1 Max | Gemma-4-31B | 4bit | oMLX | 11.4 | 66.6 | – | – | 1024 | 2026-10-06 | community | – | source |
| M1 Max | Gemma-4-31B | 4bit | oMLX | 11.4 | 66.6 | – | – | 1024 | 2026-10-06 | community | – | source |
| M1 Max | Gemma-4-31B | 4bit | oMLX | 11.1 | 66.6 | – | – | 4096 | 2026-10-06 | community | – | source |
| M1 Max | Gemma-4-31B | 4bit | oMLX | 11.1 | 66.6 | – | – | 4096 | 2026-10-06 | community | – | source |
| M1 Max | Gemma-4-31B | 4bit | oMLX | 11.1 | 66.6 | – | – | 4096 | 2026-10-06 | community | – | source |
| M5 Max | Gemma-4-31B | 4bit | oMLX | 10.8 | 379.4 | – | – | 131072 | 2026-10-11 | community | – | source |
| M4 Pro | Gemma-4-31B | 4bit | oMLX | 10.8 | 92.1 | – | – | 4096 | 2026-10-12 | community | – | source |
| M5 Ultra | Gemma-4-31B | 4bit | oMLX | 10.8 | 626.1 | – | – | 131072 | 2026-10-16 | community | – | source |
| gfx1151 | Gemma-4-31B | AWQ-4bit | vLLM | 10.5 | – | – | – | 8192 | 2026-10-13 | sourced | – | source |
| M2 Max | Gemma-4-31B | 4bit | oMLX | 10.5 | 91.0 | – | – | 1024 | 2026-10-10 | community | – | source |
| M5 Max | Gemma-4-31B | 8bit | oMLX | 10.2 | 546.8 | – | – | 32768 | 2026-10-16 | community | – | source |
| M5 Max | Gemma-4-31B | 8bit | oMLX | 10.1 | 323.0 | – | – | 131072 | 2026-10-16 | community | – | source |
| gfx1151 | Gemma-4-31B | AWQ-4bit | vLLM | 9.8 | – | – | – | 8192 | 2026-10-13 | sourced | – | source |
| M1 Max | Gemma-4-31B | 4bit | oMLX | 9.6 | 63.1 | – | – | 16384 | 2026-10-06 | community | – | source |
| M1 Max | Gemma-4-31B | 4bit | oMLX | 9.6 | 63.1 | – | – | 16384 | 2026-10-06 | community | – | source |
| M1 Max | Gemma-4-31B | 4bit | oMLX | 9.6 | 63.1 | – | – | 16384 | 2026-10-06 | community | – | source |
| M2 Max | Gemma-4-31B | 4bit | oMLX | 9.6 | 81.3 | – | – | 4096 | 2026-10-10 | community | – | source |
| gfx1151 | Gemma-4-31B | AWQ-4bit | vLLM | 9.5 | – | – | – | 128 | 2026-10-13 | sourced | – | source |
| M4 Max | Gemma-4-31B | 4bit | oMLX | 9.3 | 117.3 | – | – | 131072 | 2026-10-16 | community | – | source |
| M2 Max | Gemma-4-31B | 4bit | oMLX | 9.0 | 78.1 | – | – | 8192 | 2026-10-10 | community | – | source |
| gfx1151 | Gemma-4-31B | AWQ-4bit | vLLM | 8.5 | – | – | – | 128 | 2026-10-13 | sourced | – | source |
| M5 Max | Gemma-4-31B | 8bit | oMLX | 8.1 | 261.7 | – | – | 199680 | 2026-10-16 | community | – | source |
| M6 | Gemma-4-31B 31B | QAT | LM Studio | 8 | – | – | – | – | 2026-10 | community | – | source |
| M6 | Gemma-4-31B | 4bit | oMLX | 7.9 | 230.7 | – | – | 1024 | 2026-10-09 | community | – | source |
| M6 | Gemma-4-31B | 4bit | oMLX | 7.8 | 238.3 | – | – | 4096 | 2026-10-09 | community | – | source |
| M5 Ultra | Gemma-4-31B | 4bit | oMLX | 7.7 | 513.5 | – | – | 199680 | 2026-10-16 | community | – | source |
| gfx1151 | Gemma-4-31B | AWQ-4bit | vLLM | 6.6 | – | – | – | 128 | 2026-10-13 | sourced | contradiction | source |
| gfx1151 | Gemma-4-31B | AWQ-4bit | vLLM | 6.6 | – | – | – | 128 | 2026-10-13 | sourced | contradiction | source |
| gfx1151 | Gemma-4-31B | AWQ-4bit | vLLM | 4.6 | – | – | – | 128 | 2026-10-13 | sourced | contradiction | source |
| gfx1151 | Gemma-4-31B | AWQ-4bit | vLLM | 4.6 | – | – | – | 8192 | 2026-10-13 | sourced | contradiction outlier | source |
| K12-1 Ryzen H255 64gb | Gemma-4-31B | Q4_0 | Vulkan | 4.486488 | 73.446822 | – | – | – | 2026-09 | community | – | source |
| 2 GMKtec K12 | Gemma-4-31B | Q4_0 | llama.cpp (RPC) | 4.111661 | 18.7754 | – | – | – | 2026-09 | community | – | source |
| gfx1151 | Gemma-4-31B | AWQ-4bit | vLLM | 3.4 | – | – | – | 128 | 2026-10-13 | sourced | contradiction | source |
| gfx1151 | Gemma-4-31B | AWQ-4bit | vLLM | 2.0 | – | – | – | 8192 | 2026-10-13 | sourced | contradiction outlier | source |
| 2 GMKtec K12 | Gemma-4-31B | Q4_0 | llama.cpp (RPC) | 1.63713 | 18.288838 | – | – | – | 2026-09 | community | – | source |