community pool / model answer

Qwen3.6-27B-NVFP4

chat · unsloth/Qwen3.6-27B-NVFP4

80.5 tok/s · model median · 6 runs in the model record.

Reference rigQuantizationMedian resultBasisTTFT (ms)Peak VRAMMax contextEngines
GB10 Grace Blackwell 128GB4-bit26.4 tok/smeasured48791.1 GB262,144vllm
RTX 5090 32GB4-bit120.9 tok/sreported8229.9 GB131,072vllm
RTX PRO 6000 Blackwell 96GB4-bit138.6 tok/sreported51- GB2,048vllm

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.