Benchmark result
Tiel-Coder-35B-A3B-MLX-oQ4e-MTP on Apple M5 Max — 153.5 tok/s
Measured with oMLX 0.7.1.dev1 on October 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; the frame only adds a line that tells this page how tall it is.
How this run compares
6th fastest of 11 runs of this model on Apple M5 Max · median 153.5 tok/s.
sort
- 1200.4tok/s83.3oMLX · oQ4e · KV none
- 2200.4tok/s81.8oMLX · oQ4e · KV none
- 3181.6tok/s72.7oMLX · oQ4e · KV none
- 4157.4tok/s81.5oMLX · oQ4e · KV tq4
- 5154.4tok/s80.3oMLX · oQ4e · KV tq8
- 6153.5tok/s81.3oMLX · oQ4e · KV tq6this run
- 7126.6tok/s83.4oMLX
- 8126.2tok/s83.3oMLX
- 9124.3tok/s83.2oMLX · KV none
- 10120.9tok/s81.0oMLX
- 11120.6tok/s79.6oMLX
Tiel-Coder-35B-A3B-MLX-oQ4e-MTP on other hardware
Reproduce this run
tooloMLX 0.7.1.dev1
modelTiel-Coder-35B-A3B-MLX-oQ4e-MTP (oQ4e)
context65,536
kv cachetq6
think limitnone
samplingtemperature 0.6 · top_p 0.95 · top_k 20 · min_p 0 · presence_penalty 0 · repetition_penalty 1
clientv0.4.92+102
Set those in oMLX, then:
llm-benchmark benchmark --model "Tiel-Coder-35B-A3B-MLX-oQ4e-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.