M3 Ultra (256 GB)
13 records across 10 models (quants: 4bit, 6bit, 8bit, MXFP4, Q4).
| hardware | model | quant | backend | tok/s | pp tok/s | W | ttft s | ctx/tg | date | class | flags | source |
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| M3 Ultra (256 GB) | Devstral Small 2 24B | 4bit | MLX | 47 | – | – | – | – | – | sourced | – | source |
| M3 Ultra (256 GB) | GLM-4.5-Air | 4bit | MLX | 54 | – | – | – | – | – | sourced | – | source |
| M3 Ultra (256 GB) | GLM-4.7-Flash | 8bit | MLX | 58 | – | – | – | – | – | sourced | – | source |
| M3 Ultra (256 GB) | Qwen 3 235B-A22B | Q4 | MLX | 27 | – | – | – | – | – | sourced | – | source |
| M3 Ultra (256 GB) | qwen-3-0-6b | 4bit | MLX | 370 | – | – | – | – | – | sourced | – | source |
| M3 Ultra (256 GB) | Qwen3-Coder-Next | 4bit | MLX | 74 | – | – | – | – | – | sourced | – | source |
| M3 Ultra (256 GB) | Qwen3-Coder-Next | 6bit | MLX | 66 | – | – | – | – | – | sourced | – | source |
| M3 Ultra (256 GB) | Qwen3.5-122B-A10B | MXFP4 | MLX | 57 | – | – | – | – | – | sourced | – | source |
| M3 Ultra (256 GB) | Qwen3.5-122B-A10B | 8bit | MLX | 43 | – | – | – | – | – | sourced | – | source |
| M3 Ultra (256 GB) | Qwen3.5-27B | 4bit | MLX | 38 | – | – | – | – | – | sourced | – | source |
| M3 Ultra (256 GB) | Qwen3.5-35B-A3B | 4bit | MLX | 95 | – | – | – | – | – | sourced | – | source |
| M3 Ultra (256 GB) | Qwen3.5-35B-A3B | 8bit | MLX | 80 | – | – | – | – | – | sourced | – | source |
| M3 Ultra (256 GB) | Qwen3.5-9B | 4bit | MLX | 106 | – | – | – | – | – | sourced | – | source |
All rows cite their source. Provenance: community 40018 sourced 1085