Benchmark result
Qwen3.8-27B-oQ4-mtp on Apple M5 Max — 37.6 tok/s
Measured with oMLX 0.6.0.dev1 on August 17, 2026.
What the model built
Open full screen →The coding scenario asks for a playable game in a single HTML file. This is exactly what the model returned, unedited.
How this run compares
4th fastest of 4 runs of this model on Apple M5 Max · median 42.9 tok/s.
sort
- 147.7tok/s80.8oMLX
- 246.1tok/s87.6oMLX
- 339.7tok/s84.0oMLX
- 437.6tok/s83.2oMLXthis run
Reproduce this run
tooloMLX 0.6.0.dev1
modelQwen3.8-27B-oQ4-mtp
context65,536
clientv0.4.36+97
Set those in oMLX, then:
llm-benchmark benchmark --model "Qwen3.8-27B-oQ4-mtp" --tool "oMLX"The client prompts for context length and thinking mode, and for the KV cache dtype on Unsloth Studio. Sampling is left at the model default — the values above are what the tool reported using, not overrides the client sent.