M3 Max (40-core GPU, 128 GB)
3 records across 3 models (quants: Q4_K - Medium).
| hardware | model | quant | backend | tok/s | pp tok/s | W | ttft s | ctx/tg | date | class | flags | source |
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| M3 Max (40-core GPU, 128 GB) | llama-3-1-8b-instruct | Q4_K - Medium | llamafile | 45.792675277777775 | 587.0791163333333 | – | 2.0508701621111114 | – | – | sourced | – | source |
| M3 Max (40-core GPU, 128 GB) | llama-3-2-1b-instruct | Q4_K - Medium | llamafile | 146.33974311111112 | 3291.9091724444447 | – | 0.3584107959999999 | – | – | sourced | – | source |
| M3 Max (40-core GPU, 128 GB) | qwen-2-5-14b-instruct | Q4_K - Medium | llamafile | 25.48905755555555 | 302.37069566666673 | – | 4.070913819333334 | – | – | sourced | – | source |
All rows cite their source. Provenance: community 40018 sourced 1085