# AWS announces public SGLang support for Trainium chips

_Published Thursday, November 12, 2026 at 5:11 PM EST · AI, Infrastructure · Latest · Tier 2 — Notable_

![AWS announces public SGLang support for Trainium chips — Primary](https://d2908q01vomqb2.cloudfront.net/f1f836cb4ea6efb2a0b1b99f41ad8b103eff4b59/2026/11/12/ML-22453-featured-image.png)

Amazon announced public availability of SGLang-Neuron, a plugin that lets developers serve AI models and generate reinforcement learning training outputs on AWS Trainium chips through existing SGLang interfaces. RadixArk released it in collaboration with Annapurna Labs, with access through an AWS Deep Learning Container.

The plugin runs model execution through native PyTorch while SGLang retains its API server and request scheduling. It also supports splitting prompt processing and token generation between servers, transferring cached prompt data over Elastic Fabric Adapter. AWS's walkthrough uses a Trainium3 EC2 instance; supported models, features and validated configurations are specified in the project documentation.

## Sources

- [Artificial Intelligence](https://aws.amazon.com/blogs/machine-learning/deploy-frontier-ai-models-on-aws-trainium-with-sglang-neuron/)

---
Canonical: https://techandbusiness.org/newswire/BNN5eXqUilDSOmapRrwLov
Published: 2026-11-12T22:11:00.087Z
Story chronology: 2026-11-12T22:03:34.000Z
Retrieved: 2026-11-12T23:40:38.089Z
Publisher: Tech & Business (techandbusiness.org)