Benchmark result
qwen/qwen3.8-27b on Apple M1 Max — 14.8 tok/s
Measured with LM Studio on November 3, 2026.
Support llm-bench.io
llm-bench.io is free. If it helped you, I'd appreciate your support to cover the infrastructure costs.
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
9th fastest of 11 runs of this model on Apple M1 Max · median 16.0 tok/s.
sort
- 117.8tok/s57.0LM Studio · 4bit
- 217.4tok/s77.9LM Studio · 4bit
- 317.0tok/s86.4LM Studio · 4bit
- 416.9tok/s82.3LM Studio · 4bit
- 516.7tok/s73.9LM Studio · 4bit
- 616.0tok/s83.3LM Studio · Q4_K_M
- 715.4tok/s87.4LM Studio · Q8_0
- 815.2tok/s85.2LM Studio · 4bit
- 914.8tok/s67.1LM Studio · 4bitthis run
- 1013.9tok/s71.6LM Studio · 4bit
- 1112.5tok/s66.3LM Studio · 6bit
Reproduce this run
toolLM Studio
modelqwen/qwen3.8-27b (4bit)
context65,536
effortxhigh (model default)
clientv0.4.98+102
Set those in LM Studio, then:
llm-benchmark benchmark --model "qwen/qwen3.8-27b" --tool "LM Studio"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.