Model • LFM Diffusion • Docs • Discord
SGLang with serving support for Liquid AI's block-diffusion language
models (Lfm2ForBlockDiffusion, such as LFM2.5-350M-Diffusion).
It adds the DuoBlock algorithm: uniform-state denoising of 32-token blocks with fused Triton kernels, whole-block
CUDA graphs, commit fusion and continuous batching, behind SGLang's OpenAI-compatible server. Everything else is
upstream SGLang (2e7c85da6, August 2026). Model details, evaluation and citation are on the
LFM Diffusion page.
git clone https://github.com/Liquid4All/sglang-diffusion-lfm && cd sglang-diffusion-lfm
SGLANG_BUILD_RUST_EXTS=none pip install -e "python"SGLANG_BUILD_RUST_EXTS=none skips SGLang's optional Rust extensions, which otherwise need a Rust toolchain.
AMD (ROCm). Start from SGLang's ROCm image (mi30x for MI300X/MI325X, mi35x for MI350/MI355) and install this
repository on top of it:
docker run -it --device=/dev/kfd --device=/dev/dri --group-add video --ipc=host --shm-size 16g \
-v ~/.cache/huggingface:/root/.cache/huggingface -p 30000:30000 lmsysorg/sglang:v0.5.17-rocm720-mi30x bash
git clone https://github.com/Liquid4All/sglang-diffusion-lfm && cd sglang-diffusion-lfm
SGLANG_BUILD_RUST_EXTS=none pip install -e "python" --no-depsThe serve scripts below set SGLANG_USE_AITER=0: the ROCm images enable AITER, which slows DuoBlock by about 30%.
bash examples/dllm/lfm2_diffusion/serve_throughput.sh # batched serving, NFE 8
DECODE=decode_nfe4.yaml bash examples/dllm/lfm2_diffusion/serve_throughput.sh # fastest
bash examples/dllm/lfm2_diffusion/serve_latency.sh # one request at a timecurl http://localhost:30000/v1/chat/completions -H "Content-Type: application/json" -d '{
"model": "LiquidAI/lfm2.5-350m-diffusion-exp",
"messages": [{"role": "user", "content": "Give three tips for getting better sleep."}],
"max_tokens": 256
}'Decode presets in examples/dllm/lfm2_diffusion: decode_nfe32.yaml (highest
quality), decode_nfe8.yaml (recommended), decode_nfe4.yaml (fastest). A request can set
dllm_steps_per_block, temperature, top_p and top_k; options the block sampler does not apply (penalties,
min_p, min_new_tokens, logit_bias, structured output, logprobs) are refused with a 400.
Decode speedup over autoregressive LFM2.5-350M at batch size 1 (1024-token prompt, 1024 new tokens):
| GPU | NFE 32 | NFE 16 | NFE 8 | NFE 4 |
|---|---|---|---|---|
| H100 | 0.89× | 1.65× | 2.93× | 4.74× |
| B200 | 0.75× | 1.38× | 2.47× | 4.07× |
| MI325X | 0.78× | 1.50× | 2.81× | 4.98× |
Median of 5 machines per GPU, with the autoregressive model and the diffusion model measured back to back on the
same machine. The autoregressive baseline uses its fastest SGLang configuration (CUDA graphs and continuous decode
steps; SGLANG_USE_AITER=1 on MI325X). B200 and H100: 32 pinned CPU cores; MI325X: 14 cores. Autoregressive
decoding at batch size 1 is bound by the host CPU, so on slower CPUs the speedup is larger.
The upstream SGLang README follows.
Blog | Documentation | Roadmap | Join Slack | Weekly Dev Meeting | Slides
- [2026/07] 🔥 SGLang and Miles add day-0 support for Kimi K3 (blog).
- [2026/07] RadixArk and Google bring full SGLang features to TPUs (blog).
- [2026/07] Serving GLM5.2 NVFP4 agentic workloads with SGLang: Reaching 500 TPS in two weeks (blog).
- [2026/06] 🔥 The next generation of speculative decoding: DFlash and Spec V2 (blog).
- [2026/06] SGLang provides day-0 support for latest open models (Nemotron 3 Ultra, Nemotron 3 Super, Higgs Audio v3 TTS).
- [2026/04] 🔥 DeepSeek-V4 on Day 0: From Fast Inference to Verified RL with SGLang and Miles (blog).
- [2026/02] 🔥 Unlocking 25x Inference Performance with SGLang on NVIDIA GB300 NVL72 (blog).
- [2026/01] SGLang Diffusion accelerates video and image generation (blog).
More
- [2025/12] SGLang provides day-0 support for latest open models (MiMo-V2-Flash, Nemotron 3 Nano, Mistral Large 3, LLaDA 2.0 Diffusion LLM, MiniMax M2).
- [2025/11] SGLang Diffusion accelerates video and image generation (blog).
- [2025/10] SGLang now runs natively on TPU with the SGLang-Jax backend (blog).
- [2025/10] PyTorch Conference 2025 SGLang Talk (slide).
- [2025/10] SGLang x Nvidia SF Meetup on 10/2 (recap).
- [2025/09] Deploying DeepSeek on GB200 NVL72 with PD and Large Scale EP (Part II): 3.8x Prefill, 4.8x Decode Throughput (blog).
- [2025/09] SGLang Day 0 Support for DeepSeek-V3.2 with Sparse Attention (blog).
- [2025/08] SGLang x AMD SF Meetup on 8/22: Hands-on GPU workshop, tech talks by AMD/xAI/SGLang, and networking (Roadmap, Large-scale EP, Highlights, AITER/MoRI, Wave).
- [2025/08] SGLang provides day-0 support for OpenAI gpt-oss model (instructions)
- [2025/06] SGLang, the high-performance serving infrastructure powering trillions of tokens daily, has been awarded the third batch of the Open Source AI Grant by a16z (a16z blog).
- [2025/06] Deploying DeepSeek on GB200 NVL72 with PD and Large Scale EP (Part I): 2.7x Higher Decoding Throughput (blog).
- [2025/05] Deploying DeepSeek with PD Disaggregation and Large-scale Expert Parallelism on 96 H100 GPUs (blog).
- [2025/03] Supercharge DeepSeek-R1 Inference on AMD Instinct MI300X (AMD blog)
- [2025/03] SGLang Joins PyTorch Ecosystem: Efficient LLM Serving Engine (PyTorch blog)
- [2025/02] Unlock DeepSeek-R1 Inference Performance on AMD Instinct™ MI300X GPU (AMD blog)
- [2025/01] SGLang provides day one support for DeepSeek V3/R1 models on NVIDIA and AMD GPUs with DeepSeek-specific optimizations. (instructions, AMD blog, 10+ other companies)
- [2024/12] v0.4 Release: Zero-Overhead Batch Scheduler, Cache-Aware Load Balancer, Faster Structured Outputs (blog).
- [2024/10] The First SGLang Online Meetup (slides).
- [2024/09] v0.3 Release: 7x Faster DeepSeek MLA, 1.5x Faster torch.compile, Multi-Image/Video LLaVA-OneVision (blog).
- [2024/07] v0.2 Release: Faster Llama3 Serving with SGLang Runtime (vs. TensorRT-LLM, vLLM) (blog).
- [2024/02] SGLang enables 3x faster JSON decoding with compressed finite state machine (blog).
- [2024/01] SGLang provides up to 5x faster inference with RadixAttention (blog).
- [2024/01] SGLang powers the serving of the official LLaVA v1.6 release demo (usage).
SGLang is a high-performance serving framework for large language models and multimodal models. It is designed to deliver low-latency and high-throughput inference across a wide range of setups, from a single GPU to large distributed clusters. Its core features include:
- Fast Runtime: Provides efficient serving with RadixAttention for prefix caching, a zero-overhead CPU scheduler, prefill-decode disaggregation, speculative decoding, continuous batching, paged attention, tensor/pipeline/expert/data parallelism, structured outputs, chunked prefill, quantization (FP4/FP8/INT4/AWQ/GPTQ), and multi-LoRA batching.
- Broad Model Support: Supports a wide range of language models (Llama, Qwen, DeepSeek, Kimi, GLM, GPT, Gemma, Mistral, etc.), embedding models (e5-mistral, gte, mcdse), reward models (Skywork), and diffusion models (WAN, Qwen-Image), with easy extensibility for adding new models. Compatible with most Hugging Face models and OpenAI APIs.
- Extensive Hardware Support: Runs on NVIDIA GPUs (GB200/B300/H100/A100/Spark/5090), AMD GPUs (MI355/MI300), Intel Xeon CPUs, Google TPUs, Ascend NPUs, and more.
- Active Community: SGLang is open-source and supported by a vibrant community with widespread industry adoption, powering over 400,000 GPUs worldwide.
- RL & Post-Training Backbone: SGLang is a proven rollout backend used for training many frontier models, with native RL integrations and adoption by well-known post-training frameworks such as AReaL, Miles, slime, Tunix, verl and more.
Learn more in the release blogs: v0.2 blog, v0.3 blog, v0.4 blog, Large-scale expert parallelism, GB200 rack-scale parallelism, GB300 long context.
SGLang has been deployed at large scale, generating trillions of tokens in production each day. It is trusted and adopted by a wide range of leading enterprises and institutions, including xAI, NVIDIA, AMD, Intel, LinkedIn, Cursor, Oracle Cloud, Google Cloud, Microsoft Azure, AWS, Atlas Cloud, Voltage Park, Nebius, DataCrunch, Novita, RunPod, InnoMatrix, Modal, MIT, UCLA, the University of Washington, Stanford, UC Berkeley, Tsinghua University, Baseten, Baidu, AntGroup, Alibaba, Tencent, and other major technology organizations. As an open-source LLM inference engine, SGLang has become the de facto industry standard, with deployments running on over 400,000 GPUs worldwide. SGLang is currently hosted under the non-profit open-source organization LMSYS.
For enterprises interested in adopting or deploying SGLang at scale, including technical consulting, sponsorship opportunities, or partnership inquiries, please contact us at sglang@lmsys.org.
Long-term active SGLang contributors are eligible for coding agent sponsorship, such as Cursor, Claude Code, or OpenAI Codex. Email sglang@lmsys.org with your most important commits or pull requests.
We learned the design and reused code from the following projects: Guidance, vLLM, LightLLM, FlashInfer, Outlines, and LMQL.