community pool / model answer

Qwen3.6-27B-NVFP4

chat · nvidia/Qwen3.6-27B-NVFP4

60.1 tok/s · model median · 15 runs in the model record.

Reference rigQuantizationMedian resultBasisTTFT (ms)Peak VRAMMax contextEngines
RTX 5060 Ti 16GB ×24-bit15.9 tok/sreported140- GB2,048vllm
RTX 5090 32GB4-bit55 tok/smeasured71327.8 GB148,000llama.cpp, vllm
RTX PRO 6000 Blackwell 96GB4-bit51.3 tok/smeasured65- GB131,072tensorrt-llm, vllm
RTX PRO 6000 Blackwell 96GB ×24-bit177.7 tok/sreported46- GB2,048vllm
RTX PRO 6000 Blackwell 96GB ×44-bit168.1 tok/sreported52- 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.