Benchmark result
Qwen3.8-Flash-Next-GPTQ6Experts-Q6Attn-BF16-PLE-MLX-MTP-Q8-GPTQ4Head on Apple M4 Max — 69.1 tok/s
Measured with oMLX 0.31.3 on September 12, 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; the frame only adds a line that tells this page how tall it is.
How this run compares
1st fastest of 4 runs of this model on Apple M4 Max · median 63.3 tok/s.
sort
- 169.1tok/s81.7oMLX · KV nonethis run
- 266.3tok/s83.5oMLX · KV none
- 360.2tok/s75.0oMLX · KV none
- 459.7tok/s84.4oMLX · KV none
Reproduce this run
tooloMLX 0.31.3
modelQwen3.8-Flash-Next-GPTQ6Experts-Q6Attn-BF16-PLE-MLX-MTP-Q8-GPTQ4Head
context262,144
kv cachenone
effortlow
think limitnone
samplingtemperature 0 · top_p 1 · top_k 0 · min_p 0 · repetition_penalty 1
clientv0.4.59+97
Set those in oMLX, then:
llm-benchmark benchmark --model "Qwen3.8-Flash-Next-GPTQ6Experts-Q6Attn-BF16-PLE-MLX-MTP-Q8-GPTQ4Head" --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.