M3 Ultra (60-core GPU)
10 records across 1 models (quants: Q4_0).
| hardware | model | quant | backend | tok/s | pp tok/s | W | ttft s | ctx/tg | date | class | flags | source |
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| M3 Ultra (60-core GPU) | Qwen3.8-27B | Q4_0 | llama.cpp (Metal) | 110.0 | – | – | – | 8192 | 2026-10-02 | community | – | source |
| M3 Ultra (60-core GPU) | Qwen3.8-27B | Q4_0 | llama.cpp (Metal) | 62.6 | – | – | – | 8192 | 2026-10-02 | community | – | source |
| M3 Ultra (60-core GPU) | Qwen3.8-27B | Q4_0 | llama.cpp (Metal) | 32.45 | 317.2 | – | – | tg128 | 2026-10-02 | community | – | source |
| M3 Ultra (60-core GPU) | Qwen3.8-27B | Q4_0 | llama.cpp (Metal) | 32.36 | 317.2 | – | – | tg128 | 2026-10-02 | community | – | source |
| M3 Ultra (60-core GPU) | Qwen3.8-27B | Q4_0 | llama.cpp (Metal) | 32.1 | – | – | – | 8192 | 2026-10-02 | community | – | source |
| M3 Ultra (60-core GPU) | Qwen3.8-27B | Q4_0 | llama.cpp (Metal) | 32.1 | – | – | – | 8192 | 2026-10-02 | community | – | source |
| M3 Ultra (60-core GPU) | Qwen3.8-27B | Q4_0 | llama.cpp (Metal) | 32.0 | – | – | – | 8192 | 2026-10-02 | community | – | source |
| M3 Ultra (60-core GPU) | Qwen3.8-27B | Q4_0 | llama.cpp (Metal) | 32.0 | – | – | – | 8192 | 2026-10-02 | community | – | source |
| M3 Ultra (60-core GPU) | Qwen3.8-27B | Q4_0 | llama.cpp (Metal) | 30.2 | – | – | – | 8192 | 2026-10-02 | community | – | source |
| M3 Ultra (60-core GPU) | Qwen3.8-27B | Q4_0 | llama.cpp (Metal) | 16.8 | – | – | – | 8192 | 2026-10-02 | community | – | source |
All rows cite their source. Provenance: community 40018 sourced 1085