Pair

Qwen3.8-27B

Name: OpenWeights Terminal measured pairs
Creator: Dvidia
License: https://owterminal.com/cite

One quoted capture. Rank only lives inside a filter.

RTX PRO 6000 Blackwell Max-Q · NVFP4 · vLLM

Decode · tok/s

151.4

Prefill

—

tok/100W

60.6

同じQwen3.8-27B NVFP4重みのdecodeはTensorFold 272.8／vLLM 151.4 tok/s。RTX PRO 6000 Blackwell Max-Q・250W、1件・code greedy・256出力。下書き方式は異なります。

Quoted code-greedy decode on one PRO 6000 Max-Q. Draft method differs from the TensorFold cell in the same sentence.

Officialclaimed