M5 Max (48 GB)
4 records across 2 models (quants: 4bit, Q4_K - Medium).
| hardware | model | quant | backend | tok/s | pp tok/s | W | ttft s | ctx/tg | date | class | flags | source |
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| M5 Max (48 GB) | Qwen3.5-27B | 4bit | MLX | 31.3 | 779 | – | – | 8000 | – | sourced | – | source |
| M5 Max (48 GB) | Qwen3.5-27B | Q4_K - Medium | llama.cpp | 23.7 | 171 | – | – | 8000 | – | sourced | – | source |
| M5 Max (48 GB) | Qwen3.5-35B-A3B | 4bit | MLX | 128 | 3235 | – | – | 8000 | – | sourced | – | source |
| M5 Max (48 GB) | Qwen3.5-35B-A3B | Q4_K - Medium | llama.cpp | 89.4 | 783 | – | – | 8000 | – | sourced | – | source |
All rows cite their source. Provenance: community 40018 sourced 1085