Mia

@miaai_lab · Twitter · 2026-08-19 16:27 UTC

Qwen3.8-27B is actually GOOD now on a DGX Spark with DFlash!! 🔥 ~56 tok/s in single season (structural) - much faster in normal chat! - overall feels very smooth Get it here: https://github.com/MiaAI-Lab/Qwen3.8-27B-SGLang-DGX-Spark