community pool / model answer

Qwen3-235B-A22B-Thinking-2507

chat · Qwen/Qwen3-235B-A22B-Thinking-2507

15.6 tok/s · model median · 3 runs in the model record.

Reference rigQuantizationMedian resultBasisTTFT (ms)Peak VRAMMax contextEngines
M3 Ultra 512GB4-bit15.6 tok/smeasured3,461- GB2,048mlx

Context tested

The table above is the tested context for each exact rig and quantization cell. Empty fields remain empty rather than being filled with an estimate.