SGLang v1.0 Announced with High-Performance Inference and Scaling Updates

SGLang announced the release of version 1.0 at the SGLang Summit, introducing high-performance inference capabilities for large language models and multimodal models built for production at scale. The new version brings faster inference through zero-overhead scheduling, Rust-rewritten critical paths, and speculative decoding, alongside improved memory management and flexible scaling options. SGLang has achieved adoption across more than 330 organizations with day-zero support for 217 models. Open article →