Events

Scaling LLM Serving with Ray and SGLang. We co-hosted with Anyscale to dig into how Ray powers large-scale LLM serving with SGLang.

Ollama hosted a meetup with Google DeepMind's Gemma team in Palo Alto, and we were invited to do a live demo of deploying Gemma 4 using SGLang.

Let's Talk Gemma 4: Architecture & How to Run It on SGLang. Khoa Pham, SGLang developer, broke down Gemma 4's architecture and how to run it with SGLang. We walked through how the model works, cover what it takes to support it in SGLang, share deployment tips across GB200, GB300, and AMD.

With SGLang, Qwen, MuleRun, Novita AI, Modal, and Doubleword covering everything from frontier models to inference, agents, and production infra.

This training lab demonstrated how to optimize and scale LLM workflows with SGLang. Walked through practical performance tuning using the SGL-Cookbook, talked about profiling and bottleneck analysis, and demonstrated deep integration with Miles RL framework.

LinkedIn x SGLang Meetup: Discover the Future of LLMs for Search and Recommendation,with deep discussion into how large-scale LLM systems are built, optimized, and deployed in real-world production environments.

SGLang x RadixArk. With great food, unlimited drinks, and a relaxed evening with some of the brightest minds in model, inference, and RL infrastructure.

Featuring core members from SGLang, Qwen, Alibaba Cloud's Tair KVCache team, and partners like NVIDIA and Mooncake, covering topics from SGLang's roadmap to low-latency LLM inference optimization and KVCache storage breakthroughs.

OH on SGLang-Diffusion.CloudRipple from OpenMOSS shared how they brought MoVA model support to SGLang-Diffusion. Xingyu from Skywork AI walked through the inference optimizations powering production-ready diffusion serving.

Our core dev speaker Qiaolin Yu shared SGLang’s key priorities, upcoming initiatives, and milestones for the quarter.

SGLang x Modal Office Hours: Deploying Big MoE Models, From Zero to Serving

SGLang x OpenAnolis Meetup in Beijing. Brought together frontline engineers to break down real, production-ready system designs.

Meetup with Modal, Qwen & SGLang. The evening explored how open tools deliver competitive performance and cost for LLM inference.

On NVIDIA AI’s Dynamo Day, our core developer, Baizhou, shared firsthand insights on large-scale AI inference with the Dynamo community.

Meetup with Ant Open Source. SGLang core member shared the latest progress and roadmap, and engaged in deep discussions on inference performance, system design, and real-world deployment.

A technical deep dive into SGLang’s vision-language model capabilities, covering architecture and implementation details.

With OpenCloudOS and AMD. Featured end-to-end system optimization, low-precision performance breakthroughs, and hands-on Triton kernel development.

With AtomGit. Talked about KV cache management, inference stalls during RL/agent weight updates, new demands of GLM/Mamba/MoE models and real-world performance on Ascend hardware

With Baidu. Featured large-scale inference system optimization, distributed architectures, the latest SGLang ecosystem advances, plus an in-depth look at Baidu Baige’s inference acceleration for the DeepSeek V3 series.

At Tsinghua PACMAN Lab, we presented the SGLang Diffusion roadmap. We reused proven LLM infrastructure with a redesigned diffusion pipeline to cut inference latency by up to ~57% on H100/H200.

With AMD and Moreh AI. Showcased SGLang’s core features, roadmap, and GPU optimizations, along with hands-on exploration of open-source LLM acceleration on AMD MI300X GPUs.

New tutorial introducing SGLang Diffusion, covering high-performance image and video generation workflows with multi-GPU acceleration, CFG parallelism, and hands-on demos using CLI and Python APIs.

SGLang x Atlas Cloud NeurIPS 2025 happy hour

On NV Developer Day, our core contributor introduced SGLang, covering our core features, key performance optimizations, real-world large-scale deployment lessons, and the future roadmap.

SGLang x Volcengine x NVIDIA meetup, discussed framework updates, stability optimization, distributed KVCache acceleration, VLM and MoE inference, and efficient deployment on DGX Spark.

SGLang x Abaka AI co-hosted the EMNLP after-party, bringing together ML engineers, tech leads, and enthusiasts for conversations on AI infra, LLM agents, multimodality, and interpretability.

SGLang meetup with Meituan and AWS, brought together developers and researchers to discuss LLM inference, speculative decoding, quantization, and real-world SGLang deployments

SGLang x NVIDIA SF Meetup had incredible talks and discussions on LLM inference acceleration, distributed compute, and open infra, featuring amazing speakers from across the AI infra community.

SGLang x GOSIM Workshop, featuring deep-dive talks from core SGLang members on topics like efficient inference, PD disaggregation, and Ascend, alongside industry speakers from Qwen, NVIDIA, and Volcengine.

SGLang & AMD Meetup in San Francisco. We had a hands-on GPU workshop and talks from AMD, xAI, and the SGLang team on roadmap, MoE scaling, and inference optimization.