SGLang Office Hours (Special EP!) - 🐳 DeepSeek-V4 on Day 0: From Fast Inference to Verified RL with SGLang and Miles

​🐳 DeepSeek-V4 on Day 0: From Fast Inference to Verified RL with SGLang and Miles

​SGLang Office Hours is back!(Special EP!)

​The SGLang and Miles teams shipped Day-0 support for DeepSeek-V4, the first open-source stack to serve and train it on launch day. This session is the engineering story behind that: what the new architecture actually demanded of us, what we built to keep up, and what the numbers look like.

​Cheng Wan (SGLang Core Dev, @ChengWan17) walks us through the inference stack, from hybrid attention and ShadowRadix to the kernel and deployment work behind it.

​Yueming Yuan (Miles Core Dev, @yueming02) takes the RL side: the Day-0 verified pipeline, and the DAPO training curves.

​If you're serving DSv4 or want to see how a Day-0 launch actually gets shipped, come hang out. Bring your hardest deployment question.

Whether you're serving at scale or just getting started with the Ray integration, come with your questions.
LinkedIn Living stream 👉 SGLang Office Hour
Youtube Living stream 👉 SGLang Office Hour


Join SGLang Slack 👉 http://slack.sglang.ai/
Following us on X 👉 https://x.com/lmsysorg

​If this helps you, please consider giving us a star — it truly motivates the team. ⭐ https://github.com/sgl-project/sglang