# Yağız Çalık explains RL post-training on mixed GPU fleets

> Original title: 6 | @aiand_

- Source: LMSYS Org
- Published: 2026-11-05T22:43:28.000Z
- Bad Influence: https://badinfluence.news/items/imyt2z01z9vykz8go1x4a8k8t
- Original: https://x.com/lmsysorg/status/2118474306695160167

## Summary

Yağız Çalık shares how ai& runs RL post-training across heterogeneous GPU fleets using Miles and SGLang. He describes the inference demands that make mixed hardware a challenge.

## Post

6 | @aiand_  
RL Post-Training on Heterogeneous Infrastructure

Yağız Çalık shares how ai& runs RL across mixed GPU fleets with Miles and SGLang, and the inference demands that make heterogeneous hardware a challenge.

![](https://badinfluence.news/api/img-proxy?u=https%3A%2F%2Fpbs.twimg.com%2Fmedia%2FHWZTeXfbUAAOsHm.jpg&mode=thumb&exp=1794182400&sig=0a2f535beb71847e)