# Hugging Face TRL adds day-0 support for post-training Le Chonk on one node

> Original title: TRL has day-0 support to post-train Le Chonk on a single node 🤯

- Source: Lewis Tunstall
- Published: 2026-10-29T14:38:36.000Z
- Bad Influence: https://badinfluence.news/items/ecun36vi1devsip6mc4yfax2t
- Original: https://x.com/_lewtun/status/2115815568578097271

## Summary

Lewis Tunstall says TRL supports post-training Mistral's Le Chonk on a single node, using LoRA SFT on a 1T-parameter model with adapters on FP8 weights. He reports 32.7k tokens per step in about 8 seconds and 142 GiB per GPU on 8× B300, using expert parallelism and FSDP. The post links a training recipe on GitHub.

## Why it matters

The post gives concrete LoRA SFT figures and a linked recipe for post-training a 1T-parameter model on one node, useful for engineers planning similar fine-tuning.

## Post

TRL has day-0 support to post-train Le Chonk on a single node 🤯

LoRA SFT on a 1T-parameter model, training adapters on top of FP8 weights:

⚡ 32.7k tokens per step in ~8 seconds
💾 142 GiB per GPU on 8× B300
🔀 Expert parallelism + FSDP

Recipe: https://github.com/huggingface/trl/blob/ea936f801deeea713b032b7d4d87dfabd58a52d1/examples/sft_mistral_large_4/sft_mistral_large_4.py

Open weights = open training after all ;)

## Quoted post @MistralAI

> Today, we release the weights of a new and improved Le Chonk, advancing the open frontier for agentic coding and cybersecurity.
> 
> 💻 Built to build: Le Chonk ranks #1 on @cognition’s FrontierCode among open models, and sets a new benchmark for Pareto efficiency on @artificialanalysis.
> 
> 🛡 Built to defend: State-of-the-art in cyberdefense, put through the paces by the world's leading cyber vendors, including @CrowdStrike, @cloudflare, @rubrik, @zscaler, and others.
> 
> Built for sovereignty: Make it yours. Download the weights on @huggingface , customize it with Mistral Forge, and deploy it wherever you need it.
> 
> Read more from our team and customers 👇:

https://x.com/MistralAI/status/2115811242556137820