community pool / model answer

Qwen3.6-35B-A3B-FP8

chat · Qwen/Qwen3.6-35B-A3B-FP8

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

Reference rigQuantizationMedian resultBasisTTFT (ms)Peak VRAMMax contextEngines
CMP 170HX 64GB8-bit288.1 tok/sreported5555.8 GB65,536vllm
GB10 Grace Blackwell 121GB8-bit54.1 tok/sreported-- GB128,000vllm
RTX 3090 24GB ×28-bit140 tok/sreported14922.2 GB32,768vllm
RTX PRO 6000 Blackwell 96GB8-bit253.7 tok/sreported22938.4 GB4,096vllm

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.