---
format: "aidr-story-markdown/v1"
id: "0db7b6a64d84e6230ade0db24221a333c20a4c1d9ed4b400f69248a768fcc501"
canonical_url: "https://aidr.today/0db7b6a6?lang=en"
title: "SGLang and vLLM Match DeepSeek's DeepInfer Engine With Full Production Serving Stacks"
lang: "en"
requested_lang: "en"
available_langs: ["en","vi"]
translation_fallback: null
fallback_fields: []
published_at: "2026-10-24T19:23:01.000Z"
category: "Frameworks"
topics: ["deepseek","inference","framework"]
source_urls: ["https://marketbrief.now/ai/update-sglang-and-vllm-match-deepseeks-deepinfer-engine-with-full-produc-a1636bba","https://x.com/vllm_project/status/2114054277379641770","https://x.com/vllm_project/status/2114054289488478256","https://x.com/sgl_project/status/2114071115605877043","https://x.com/sgl_project/status/2114071118940360743","https://x.com/natolambert/status/2113685620941885732","https://x.com/SemiAnalysis_/status/2112950867301159252","https://x.com/swyx/status/2112587397380943953"]
summary: "Open source inference frameworks SGLang and vLLM integrated DeepSeek's kernels and optimizations to match or exceed the performance of DeepInfer on the 552B parameter DeepSeek-V4.1-Flash model. SGLang achieved up to 1.04 times DeepInfer's prefill throughput on four Nvidia GB300 or B300 GPUs and up to 7.4% higher decode throughput across 64 GB300 chips. vLLM matched DeepInfer's decode curve across batch sizes 8 to 256 using half the expert parallelism and reached 954,000 tokens per second per GPU on the SemiAnalysis AgentX benchmark. Both frameworks provide the production serving infrastructure that DeepInfer omitted when it was open sourced under an Apache 2.0 license on Oct. 20."
---

# SGLang and vLLM Match DeepSeek's DeepInfer Engine With Full Production Serving Stacks

> [Open the canonical story](<https://aidr.today/0db7b6a6?lang=en>)

**Published:** 2026-10-24T19:23:01.000Z
**Category:** Frameworks
**Topics:** deepseek, inference, framework

## Summary

Open source inference frameworks SGLang and vLLM integrated DeepSeek's kernels and optimizations to match or exceed the performance of DeepInfer on the 552B parameter DeepSeek\-V4\.1\-Flash model\. SGLang achieved up to 1\.04 times DeepInfer's prefill throughput on four Nvidia GB300 or B300 GPUs and up to 7\.4% higher decode throughput across 64 GB300 chips\. vLLM matched DeepInfer's decode curve across batch sizes 8 to 256 using half the expert parallelism and reached 954,000 tokens per second per GPU on the SemiAnalysis AgentX benchmark\. Both frameworks provide the production serving infrastructure that DeepInfer omitted when it was open sourced under an Apache 2\.0 license on Oct\. 20\.

## Sources

- [Story source](<https://marketbrief.now/ai/update-sglang-and-vllm-match-deepseeks-deepinfer-engine-with-full-produc-a1636bba>)
- [Story source](<https://x.com/vllm_project/status/2114054277379641770>)
- [Story source](<https://x.com/vllm_project/status/2114054289488478256>)
- [Story source](<https://x.com/sgl_project/status/2114071115605877043>)
- [Supporting source](<https://x.com/sgl_project/status/2114071118940360743>)
- [Supporting source](<https://x.com/natolambert/status/2113685620941885732>)
- [Supporting source](<https://x.com/SemiAnalysis_/status/2112950867301159252>)
- [Supporting source](<https://x.com/swyx/status/2112587397380943953>)