NVIDIA、AMD、Intel、Google、Meta 等组建 SGLang 推理引擎联盟
9 家公司一起给 SGLang 这个开源推理引擎写代码、跑测试,AMD 和 Intel 加进来还意味着不只 NVIDIA 卡能用。自己搭推理服务的开发者可以直接关注这个项目的后续版本。
SGLang 是一个开源推理引擎,NVIDIA、AMD、Intel、Google、Meta、Thinking Machines Lab、Cloudflare、Modal 和 RadixArk 等 9 家公司宣布组建联合体来共同改进它。联盟将为 SGLang 投入工程开发资源和测试资源。这些公司还会把各自大规模运行 AI 推理的实际经验沉淀到项目中。
对开发者来说,这意味着 SGLang 在不同硬件(NVIDIA、AMD、Intel 芯片)上的兼容性和稳定性会获得多家厂商的直接支持,社区版推理引擎有望更快吸收生产环境中的规模化部署经验。开源推理引擎由多家头部公司共同维护的这种组织形式,对依赖自建推理服务的企业团队是一个值得关注的信号。
SGLang is an open-source inference engine, and it's about to get a whole lot better:
NVIDIA, AMD, Intel, Google, Meta, Thinking Machines Lab, Cloudflare, Modal, and RadixArk are joining a consortium to improve it.
This is really good for the community!
SGLang will be getting:
• Engineering work • Testing resources • Lessons on how to run AI at scale from these companies
This report is categorized under AITOP's "Industry" desk, focusing on US frontier AI labs and Silicon Valley technology leaders. Within our cross-regional analytical framework, it directly maps to AI developer toolchains, coding agents, and engineering workflows. Key entities and technical topics under continuous tracking include SGLang, inference engines, NVIDIA, open-source models.
Source attribution: curated from Ate-a-Pi (@svpino).