SGLang
In plain English
SGLang is open-source software for serving language models quickly, with scheduling and optimization features for complex AI workloads.
Technical definition
SGLang is an open-source serving engine and language-model programming system optimized for high-performance LLM and multimodal inference.
Engineering details
The serving runtime includes continuous batching, prefix-aware scheduling, distributed parallelism, speculative decoding, and multiple attention and MoE kernel backends across NVIDIA and AMD GPUs.
Why it matters
SGLang releases and model-specific kernel work can change throughput on the same hardware. Scheduler overhead matters at low concurrency, while attention, MoE, and communication kernels dominate other regions.
How to read it in InferenceX
InferenceX continuously reruns pinned SGLang recipes. Version-to-version curves show where a change affects performance across the operating range and reveal regressions or gains hidden by one peak point.
Source material
See the concept in real benchmarks
SGLang 0.5.6 on B200 DeepSeek R1 FP4: Up to 1.8x at Low Concurrency
Piecewise CUDA graphs for DeepSeek V3, a unified event loop, and JIT kernels push 8k/1k throughput from 508 to 907 tok/s/GPU on the same 16 GPU B200 pool
B200 NVFP4 vs H200 FP8 on GLM-5: Up to 3.65x Better Performance per Dollar with SGLang MTP
Both SKUs run SGLang EAGLE MTP; the Blackwell generation lifts perf/$ by ~1.2x at the peak and the NVIDIA GLM-5-NVFP4 checkpoint on FlashInfer TRT-LLM sparse MLA stacks another ~2.4–3.0x on 8K/1K
MI355X DeepSeek-V4-Pro on SGLang: 110.5x Throughput per GPU in 26 Days
The amd/deepseek_v4 side branch shipped TileLang attention indexer, Triton sparse MLA, fused RoPE/Hadamard, FlyDSL MoE, and FP4 weights across 31 performance optimizations PRs — lifting first-light 20 tok/s/GPU at 2.4 tok/s/user into 2,256 tok/s/GPU at 9.4 tok/s/user on 8K/1K, with both throughput and interactivity climbing together
AMD MI355X GLM-5 Inference: Up to 40% Cheaper per Million Tokens than B200 on SGLang FP8
14 weeks after GLM-5 launched, AMD landed both MTP and non-MTP SGLang FP8 recipes on MI355X — fused MLA + FP8 KV cache via TileLang flips the single-node FP8 cost curve in AMD favor across most of the performance Pareto
AMD MI355X Qwen3.5 397B-A17B Inference: Up to 19x Throughput per GPU in 3 Months on SGLang FP8
From v0.5.8 (Feb) → v0.5.10rc0 (Apr) → v0.5.12 (May), three AITER kernel landings on MI355X plus a TP=8 → TP=2/TP=4 retune push Qwen3.5 8k/1k peak from 1.3k to 6.4k tok/s/GPU and extend the curve out to 75 tok/s/user