Benchmark result
Ornith-1.5-35B-A3B-oQ8e-mtp on Apple M5 Max — 108.0 tok/s
Measured with oMLX 0.7.0.dev4 on September 24, 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
4th fastest of 11 runs of this model on Apple M5 Max · median 95.9 tok/s.
sort
- 1129.1tok/s84.7oMLX · oQ8e · KV none
- 2128.2tok/s83.9oMLX · oQ8e · KV none
- 3121.5tok/s85.1oMLX · oQ8e · KV none
- 4108.0tok/s85.8oMLX · oQ8e · KV nonethis run
- 598.3tok/s84.4oMLX
- 695.9tok/s84.4oMLX · oQ8e · KV none
- 795.6tok/s78.9oMLX
- 894.7tok/s85.1oMLX
- 992.7tok/s85.2oMLX
- 1092.1tok/s84.5oMLX
- 1191.8tok/s85.0oMLX · KV none
Reproduce this run
tooloMLX 0.7.0.dev4
modelOrnith-1.5-35B-A3B-oQ8e-mtp (oQ8e)
context262,144
kv cachenone
think limitnone
samplingtemperature 1 · top_p 0.95 · top_k 0 · repetition_penalty 1
clientv0.4.80+98
Set those in oMLX, then:
llm-benchmark benchmark --model "Ornith-1.5-35B-A3B-oQ8e-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.