# NVIDIA explains how Dynamo works with SGLang for agentic inference

> Original title: Agentic inference and RL rollouts need long contexts, cache reuse and fast weight updates.

- Source: LMSYS Org
- Published: 2026-11-09T23:15:37.000Z
- Bad Influence: https://badinfluence.news/items/i068552y2sjn0suw8198cc2vd
- Original: https://x.com/lmsysorg/status/2119931946763800733

## Summary

NVIDIA says Dynamo integrates with SGLang to support KV-aware routing, disaggregated serving, and trainer-to-inference weight transfer. The post targets agentic inference and RL rollouts, which need long contexts, cache reuse, and fast weight updates.

## Post

Agentic inference and RL rollouts need long contexts, cache reuse and fast weight updates.

Ria Jain (@nvidia) shares how Dynamo works with SGLang on KV-aware routing, disaggregated serving and trainer-to-inference weight transfer.

![](https://badinfluence.news/api/img-proxy?u=https%3A%2F%2Fpbs.twimg.com%2Fmedia%2FHWuBiD_bUAA-1Xv.jpg&mode=thumb&exp=1794528000&sig=78358932ae9a1479)