M3 Max (40-core GPU, 48 GB)
2 records across 2 models (quants: Q4_0, 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, 48 GB) | Llama 2 7B | Q4_0 | llama.cpp | 65.85 | 690.99 | – | – | 512 | – | sourced | – | source |
| M3 Max (40-core GPU, 48 GB) | llama-3-2-1b-instruct | Q4_K - Medium | llamafile | 148.95889516666665 | 3399.1241412222225 | – | 0.34880203711111113 | – | – | sourced | – | source |
All rows cite their source. Provenance: community 40018 sourced 1085