M1 Ultra (64 GB)
9 records across 5 models (quants: 4bit, 8bit, Q4_K - Medium, Q8, oQ4e, oQ6e).
| hardware | model | quant | backend | tok/s | pp tok/s | W | ttft s | ctx/tg | date | class | flags | source |
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| M1 Ultra (64 GB) | CyberTiel-Coder-35B | oQ6e | MLX | 116 | 1890 | – | – | – | 2026-10-07 | community | – | source |
| M1 Ultra (64 GB) | Devstral Small 2 24B | 4bit | MLX | 29.68 | – | – | 6.63 | – | – | sourced | – | source |
| M1 Ultra (64 GB) | Devstral Small 2 24B | Q4_K - Medium | llama.cpp | 25.3 | – | – | 5.89 | – | – | sourced | – | source |
| M1 Ultra (64 GB) | Devstral Small 2 24B | Q8 | llama.cpp | 23.37 | – | – | 5.66 | – | – | sourced | – | source |
| M1 Ultra (64 GB) | Devstral Small 2 24B | 8bit | MLX | 22.32 | – | – | 7.57 | – | – | sourced | – | source |
| M1 Ultra (64 GB) | Llama 3.3 70B | 4bit | LM Studio | 12.55 | – | – | 49.8 | 3991 | – | sourced | – | source |
| M1 Ultra (64 GB) | MiniCPM-V-4.6 | 8bit | MLX | 144 | – | – | – | – | 2026-10-07 | community | – | source |
| M1 Ultra (64 GB) | Qwen3.8-27B-Huihui-Abliterated | oQ4e | MLX | 49 | 280 | – | – | – | 2026-10-07 | community | – | source |
| M1 Ultra (64 GB) | Qwen3.8-27B-Huihui-Abliterated | oQ4e | MLX | 38 | – | – | – | 16384 | 2026-10-07 | community | – | source |
All rows cite their source. Provenance: community 40018 sourced 1085