Source: SGLang documentation (home)

Official SGLang docs landing page: positioning, hardware support and adoption claims as of late 2026.

Contents
  • What: official documentation (read 2026-11-14)
  • Author: SGLang project (LMSYS)
  • Date: 2026-11-14 (accessed)
  • Original: SGLang documentation (home)

Summary

The landing page positions SGLang as a high-performance serving framework built around RadixAttention and prefix caching, scaling from one GPU to large clusters.

Key claims

  • Supports NVIDIA, AMD, Intel Xeon, Google TPU, Ascend NPU and Moore Threads accelerators.
  • OpenAI- and Hugging Face-compatible APIs; models such as Llama, Qwen, DeepSeek.
  • Self-reported: trillions of tokens per day and more than 400,000 GPUs in deployment; integrated with RL training frameworks.

Feeds

LLM inference serving

Suggest changes →