# Blackwell-native MXFP8 and NVFP4 RL with SGLang and Miles lessons

> Original title: 4 | @humansand

- Source: LMSYS Org
- Published: 2026-11-05T22:43:27.000Z
- Bad Influence: https://badinfluence.news/items/bsayylsdhw5q0164u27paycc5
- Original: https://x.com/lmsysorg/status/2118474301288616014

## Summary

Ziang Li from humansand shares lessons on trainer-sampler consistency, gradient stability, and precision control for bringing low-precision reinforcement learning with MXFP8 and NVFP4 on Blackwell into production. The post, shared by LMSYS Org, covers production RL built on SGLang and Miles.

## Post

4 | @humansand 
Production Blackwell-Native MXFP8 & NVFP4 RL with SGLang and Miles

Ziang Li shares lessons on trainer-sampler consistency, gradient stability, and precision control when bringing low-precision RL to production.

![](https://badinfluence.news/api/img-proxy?u=https%3A%2F%2Fpbs.twimg.com%2Fmedia%2FHWZTGDLa8AEhlsv.jpg&mode=thumb&exp=1794182400&sig=8a35aba01b15a896)