LMSYS Launches SGLang Inference Consortium With Nvidia, AMD, Intel, Google and Meta
At the SGLang Summit in San Francisco, LMSYS Org formed the SGLang Consortium alongside major silicon and cloud providers to standardise open-source AI inference.

By Marcus Vance, Semiconductors & Hardware
, updated 1 min read
LMSYS Org has formally launched the SGLang Consortium at the SGLang Summit in San Francisco, bringing together Nvidia, AMD, Intel, Google, and Meta to coordinate open-source AI inference software and hardware support.
Standardising high-throughput model serving
SGLang has become one of the primary open-source engines for serving large language and multimodal models in production, competing with vLLM and vendor-specific runtimes. The consortium aims to align kernel optimisations, structured generation features, and multi-node scheduling across competing GPU and accelerator architectures.
Why hardware vendors are backing one runtime
For AMD, Intel, and Google Cloud, shared upstream support in SGLang lowers the engineering cost of running frontier open-weight models efficiently outside single-vendor stacks, while Nvidia gains tighter co-design on new tensor and interconnect features.
Sources
- 1Business Wire: SGLang Inference Consortium launchbusinesswire.com