community pool / model answer

Qwen3-Coder-30B-A3B-Instruct-4bit

code · mlx-community/Qwen3-Coder-30B-A3B-Instruct-4bit

120.9 tok/s · model median · 2 runs in the model record.

Reference rigQuantizationMedian resultBasisTTFT (ms)Peak VRAMMax contextEngines
M5 Max 128GB4-bit140.7 tok/sreported-17.5 GB2,048mlx
M5 Pro 64GB4-bit101.1 tok/sreported22217.5 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.