SGLang Logo
GitHub

Blog

Releases, benchmarks, and deep dives from the SGLang team.

M3.1 Flash×MiniMaxSGLang
October 22, 2026

SGLang and Miles Add Day-0 Support for MiniMax M3.1 Flash

SGLang & Miles Team
Day-0 Support
Fixed poolsUnusedShared bufferRecurrent stateKV cache
October 22, 2026

Unified Memory in SGLang: 1.9× as many prefixes retained, 37.9% less time for returning requests

SGLang Team
SGLang Deep Dive
full-duplex & interactive AI
October 20, 2026

Unified Runtime for Full-Duplex Interactive Intelligence with SGLang Omni

SGLang Omni Team
SGLang Deep Dive
GPU HBM · L1CPU DRAM · L2External storage · L3
October 19, 2026

HiCache in SGLang: hybrid-model caching with up to 15.5× input throughput

SGLang Team
SGLang Deep Dive
Step 5 Preview×StepFunSGLang
October 14, 2026

Step 5 Preview: Day-0 Inference and RL Support with SGLang and Miles

SGLang & Miles Team
Day-0 Support
EncodersThinkerTalkerVocoder
October 13, 2026

SGLang Omni: Serving Multi-Stage Omni Models with Heterogeneous Stage Scheduling

SGLang Omni Team
SGLang Deep Dive
Shared prefixFULLSWAMAMBAKV pagesWindowCheckpoint
September 23, 2026

Unified Radix Cache in SGLang: one tree for hybrid model prefix caching

SGLang Team
SGLang Deep Dive
×RLinfSGLang
September 20, 2026

RLinf x SGLang: Integrating Cosmos3 from Fine-Tuning to Efficient Parallel Evaluation

RLinf Team & SGLang Team
SGLang Deep Dive
DeepSeek V4.1 Flash on SGLang, 873 tokens per second
September 10, 2026

DeepSeek-V4.1 Flash on SGLang: from 35 to 873 tokens/s

SGLang Team
SGLang Deep Dive
V4.1×DeepSeekSGLang
September 9, 2026

SGLang and Miles Add Day-0 Support for DeepSeek-V4.1

SGLang & Miles Team
Day-0 Support
EagerLaunch each kernelCPUGPUBCGReplay graph segmentsCPUGPUGraphEager opIdle
August 27, 2026

Breakable CUDA Graph in SGLang: 5× faster graph builds, 1.93× faster prefill

SGLang Team & Meta Team
SGLang Deep Dive
Kimi K3×Moonshot AISGLang
August 19, 2026

Kimi K3: Architecture and SGLang Day-0 Support

SGLang Team & Moonshot AI Team
Day-0 SupportSGLang Deep Dive