TechAILatest.newsTrackers
Open Source AI / Open tools

SGLang Consortium launches with NVIDIA, AMD, Intel, Google and Meta as founding members

LMSYS Org launched the SGLang Consortium on 12 November 2026 with eight founding members who help set the roadmap of the open-source SGLang inference engine.

By Emre A.Published

Abstract illustration of a branching code repositoryTech AI Latest
Illustration of a branching code repository, made by Tech AI Latest for this story. Illustration: Tech AI Latest (Original illustration)

LMSYS Org, the nonprofit behind the open-source SGLang inference engine, launched the SGLang Consortium on 12 November 2026 at the first SGLang Summit in San Francisco. The eight founding members are NVIDIA, AMD, Intel, Google, Meta, Cloudflare, Modal and RadixArk. They will work with SGLang's maintainers on the roadmap, while the maintainers keep control of technical decisions and which code is merged. No funding amount or performance target was announced.

Key facts

  • Announced: 12 November 2026 by LMSYS Org, at the opening keynote of the first SGLang Summit in San Francisco (LMSYS Org)
  • Founding members: NVIDIA, AMD, Intel, Google, Meta, Cloudflare, Modal and RadixArk
  • Governance: members help shape the roadmap; maintainers keep control of technical decisions and code adoption
  • Focus areas: frontier model support, reasoning and multimodal workloads, distributed inference, hardware optimization
  • Not stated: membership fees, a budget, or performance targets for the consortium

What LMSYS announced

LMSYS Org, the nonprofit research group that maintains SGLang, announced the SGLang Consortium on 12 November 2026 in the opening keynote of the first SGLang Summit, held in San Francisco on 12 and 13 November. The announcement was also distributed as a Business Wire press release. The LMSYS site cannot be opened from our newsroom tools; the details below are those in the release as republished by Yahoo Finance and HPCwire, and in reports by SDxCentral and Igor's Lab that agree on them.

The consortium has eight founding members: chip makers NVIDIA, AMD and Intel, Google, Meta, Cloudflare, the GPU cloud platform Modal, and RadixArk. According to the release, members will work with SGLang's maintainers to help shape the project's roadmap, and the consortium gives them a forum to discuss inference trends and turn shared priorities into joint engineering work. The named areas are support for frontier models, reasoning and multimodal workloads, distributed inference across many machines, and optimization for different hardware.

The maintainers keep authority over the code. The release says they retain control of technical decisions and of which contributions are adopted, so membership does not buy a vote on merges. LMSYS did not publish membership fees, a budget, or any performance target, and coverage of the launch notes that no benchmark result is tied to the consortium itself.

Why rival chip makers are joining

SGLang is one of the most widely used open-source engines for serving large language models, alongside projects such as vLLM. It grew out of a 2023 research paper by a team led by Lianmin Zheng and Ying Sheng, which introduced RadixAttention, a way to reuse cached computation across requests that share a prefix. The project's own site says SGLang runs on more than 400,000 GPUs in production and generates trillions of tokens a day, and lists xAI, Google, Microsoft, NVIDIA, Oracle, AMD and LinkedIn among the organizations that use it. Those are the project's figures.

For hardware vendors, an inference engine decides how quickly a new chip can serve the open-weight models that customers actually run. NVIDIA, AMD and Intel compete directly for the same data centre budgets, and each wants its accelerators to be supported well and early in the software layer that labs and cloud providers already use. A shared roadmap forum lets each of them push for its own hardware paths without forking the project. Commentary on the launch has been divided: some outlets call the line-up remarkable for rivals, others describe membership as defensive table stakes for chip makers.

For users, the practical question is whether coordinated work reaches releases faster, for example support for new model architectures on non-NVIDIA hardware, or for multi-node serving of very large models. LMSYS has not given dates for any of that.

Security and what to watch

The launch comes days after SGLang was in the news for a different reason. NEAR AI said a remote code execution flaw in SGLang was used to see confidential inference requests on its platform on 8 November, and CERT/CC had earlier published a note on six SGLang vulnerabilities, including unauthenticated remote code execution. The consortium release does not mention security work among its focus areas; whether members fund audits or a security response process is an open question.

Other open inference work on the site includes Hugging Face running Transformers natively on Google TPUs. Signs to watch next are the first roadmap published under the consortium, whether more members join after the founding group, and which hardware back ends gain first-class support in the coming SGLang releases. More coverage of open models and tools is on the Open Source AI hub.

What changed

BeforeSGLang was developed by LMSYS and community maintainers with contributions from companies but no formal industry body
AfterEight founding members, including NVIDIA, AMD, Intel, Google and Meta, now have a formal role in shaping SGLang's roadmap, while maintainers keep control of the code
Who it affectsTeams that serve open-weight models with SGLang, inference providers, and chip makers competing for inference workloads

Questions people ask

Who are the members of the SGLang Consortium?
The eight founding members are NVIDIA, AMD, Intel, Google, Meta, Cloudflare, Modal and RadixArk, according to the LMSYS Org announcement of 12 November 2026.
Do consortium members control SGLang?
No. Members help shape the roadmap, but the announcement says SGLang's maintainers keep control of technical decisions and which code is adopted.
Does the consortium change SGLang's licence?
The announcement does not mention any licence change. SGLang remains an open-source project on GitHub.

Sources

  1. LMSYS Org Launches SGLang Consortium with NVIDIA, AMD, Intel, Google, Meta, Cloudflare, Modal and RadixArk to Advance Open-Source AI Inference (LMSYS Org, 12 Nov 2026)
  2. LMSYS Org Launches SGLang Consortium (Business Wire release) (Yahoo Finance, 12 Nov 2026)
  3. LMSYS Org Forms SGLang Consortium with NVIDIA, AMD, Google and Meta (HPCwire, 13 Nov 2026)
  4. Nvidia gets back on the horse with AI inference pledge (SDxCentral, 13 Nov 2026)
  5. SGLang: Efficient Execution of Structured Language Model Programs (arXiv, 12 Dec 2023)

Reference pages

Spotted an error? See the corrections policy and tell the editor. Facts in this story come from the sources listed above; vendor claims are attributed.

Comments

Loading comments.

No links, no insults; comments that break the rules are not published. 0/1000

Related stories

All open source ai