Qwen3.8-27B FP8 performance on one RTX PRO 6000
Four coding requests reached a median 94.3 tok/s each on one RTX PRO 6000. Qwen3.8-27B FP8 setup, latency results, and benchmark limits.
The journal
Tag: LLM inference
Read the story : Qwen3.8-27B FP8 performance on one RTX PRO 6000