🔥The 9th Round of Easy Loan, Earn $40 Reward is in progress❗️ ⏰ Promotion Period: January 15th - Feburary 15th, 2025 👉 Register now and check more details at gate.io/campaigns/358

x.com/i/article/1861…

We will present latest updates of verl at #ICLR2025: - recent RL recipes (DAPO, etc) - RL with tool calling & multi-turn - full sglang integration (with LMSYS Org ) - large scale optimizations, and many more Come join us!

Excited to co-host the "Frontiers of Generative AI" afterparty with Abaka AI at #ICLR2025! 🚀 Come connect with fellow researchers and enthusiasts from LMSYS Org and beyond. 🗓️ When: April 26th (Sat), 18:30 - 21:30 🗺️ Where: 10-min from Singapore EXPO (Register for exact details)

The SkyPilot team's benchmark for SGLang and vLLM shows that SGLang consistently outperforms with a 30% performance advantage. We will continue to optimize. Cheers!

Introducing Grok Vision, multilingual audio, and realtime search in Voice Mode. Available now. Grok habla español Grok parle français Grok Türkçe konuşuyor グロクは日本語を話す ग्रोक हिंदी बोलता है

Around this time 2 years ago, the community helped us launch the very first Arena leaderboard! Today we’re publishing a blog to celebrate everything we’ve built together on LMArena! 🥳👏 Highlights: ☑️ 3M+ community votes 🤖 400+ models ranked across text, vision,

LMSYS Org thanks to the support of SGL! They have been dedicated to helping us deploy the large MoE models and improve the inference efficiency. 🌊

📷  Thrilled to share my latest blog post  (co-authored with Qingquan (QQ) Song ) on our open source work integrating the Flash Attention Backend end-to-end in SGLang—now enabled by default in the latest version! 😎 Dive into the details here: hebiao064.github.io/fa3-attn-backe… #LLM #SGLang

🚨 New Challenger: GROK joins the Game Arena Benchmark! We evaluated Grok3-mini-beta: thinkining on four games: 🧩 2048 | 🧱 Sokoban | 🍬 Candy Crush | 🎮 Phoenix Wright With fast progress, it’s already comparable to top models like OpenAI’s O1, previous O3-mini, and

🚀 Breaking: SGLang provides the first open-source implementation to serve DeepSeek V3/R1 models with large-scale expert parallelism and prefill-decode disaggregation on 96 GPUs. It nearly matches the throughput reported by the official DeepSeek blog, achieving 52.3K input

MLSys 2025 is coming up! Want to meet the developers behind FlashInfer, XGrammar, and SGLang LMSYS Org in person? Join us for the Happy Hour on May 12—we’d love to see you there! lu.ma/dl99yjoe

E2E guide to self-host multi-node Llama 4 with one command, using SkyPilot, SGLang (LMSYS Org), and `llm` (Simon Willison). ▸ Spin up SGLang on 8x H100 (single and multi-node) ▸ Perf numbers included: token/s, TTFT, ITL ▸ Production-readiness: auth, HTTPS ▸ Tooling: easy

We are hosting a happy hour with LMSYS Org at #mlsys2025! Join us for engaging talks on SGLang, the structured generation library XGrammar, and the high-performance kernel library FlashInfer. Enjoy great food, lively discussions, and connect with the community! Click to join 👉

Please offer galaxy brain questions and answers. This is for Grok training. docs.google.com/forms/d/e/1FAI…

I’ll be joining my Baseten colleague Philip Kiely at the AI Engineer World’s Fair AI Engineer in San Francisco, June 3–5, to Introduce LLM serving with SGLang LMSYS Org. We’d love for you to stop by and exchange ideas in person!🤗

We’re thrilled that FlashInfer won a Best Paper Award at MLSys 2025! 🎉 This wouldn’t have been possible without the community — huge thanks to LMSYS Org’s sglang for deep co-design (which is crtical for inference kernel evolution) and stress-testing over the years, and to

Thrilled to partner with LMSYS Org! 🚀 Our high-performance GPU cloud is powering their game-changing inference engine, revolutionizing LLM deployment with breakthrough RL framework and multi-LLM serving capabilities. Together, we're bringing lightning-fast AI inference to

(1/5) We’ve never enjoyed watching people chop Llamas into tiny pieces. So, we’re excited to be releasing our Low-Latency-Llama Megakernel! We run the whole forward pass in single kernel. Megakernels are faster & more humane. Here’s how to treat your Llamas ethically: (Joint

We sincerely appreciate NVIDIA’s generous sponsorship. It is truly regrettable that the SGLang team was unable to attend in person and witness the physical hardware firsthand. Our special thanks go to InnoMatrix for their invaluable assistance in managing everything seamlessly.

Blackwell-specific optimizations are cooking! 🚀