M3 Max (30 GPU)
3 records across 1 models (quants: F16, Q4_0, Q8_0).
| hardware | model | quant | backend | tok/s | pp tok/s | W | ttft s | ctx/tg | date | class | flags | source |
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| M3 Max (30 GPU) | llama 7B | Q4_0 | llama.cpp (Metal) | 56.58 | 567.59 | – | – | – | 2023-11-22 | community | – | source |
| M3 Max (30 GPU) | llama 7B | Q8_0 | llama.cpp (Metal) | 34.3 | 566.4 | – | – | – | 2023-11-22 | community | – | source |
| M3 Max (30 GPU) | llama 7B | F16 | llama.cpp (Metal) | 19.54 | 589.41 | – | – | – | 2023-11-22 | community | – | source |
All rows cite their source. Provenance: community 40018 sourced 1085