SGLang(@sgl_project):We're adding a few Qwen3.8-Flash-Next recipes to the SGLang cookbook, each verified on the hardware: - 1x RTX PRO 6000, PLE table in system RAM - 1x DGX Spark, PLE table on local NVMe - 2x DGX Spark, TP=2 - NVFP4 support for both the RadixArk export and @NVIDIAAI's ModelOpt export Cookbook: https://t.co/XtdY5WidJz Thanks to everyone who has been testing and sharing feedback. We've been following the community reports closely and fixed a number of issues that came with this new architecture. More performance work is on the way, and we want to be ready for what comes next from the Qwen family!

SGLang

@sgl_project

Run LLMs fast at any scale 🔗 Join our community For AI tech blogs & deep-dives 👉 @lmsysorg

参加 May 2025

53 フォロー中 9.5K ファン

SGLang@sgl_project

2026.09.07 23:01

We're adding a few Qwen3.8-Flash-Next recipes to the SGLang cookbook, each verified on the hardware: - 1x RTX PRO 6000, PLE table in system RAM - 1x DGX Spark, PLE table on local NVMe - 2x DGX Spark, TP=2 - NVFP4 support for both the RadixArk export and @NVIDIAAI's ModelOpt export Cookbook: Thanks to everyone who has been testing and sharing feedback. We've been following the community reports closely and fixed a number of issues that came with this new architecture. More performance work is on the way, and we want to be ready for what comes next from the Qwen family!