A training framework born from real-world, large-scale production workloads β and shared back with the community
LoongForge is an open-source training framework developed by the Baidu AI Cloud Baige team, built to deliver faster training for mainstream LLMs, VLMs, diffusion, and embodied models β and thereby significantly reduce cost
LoongForge was open-sourced from AIAK-Training-LLM, a training acceleration suite delivered to enterprise customers on Baidu AI Cloud, after years of hardening under real production workloads:
It now joins the Baige Loong open-source series β named after the traditional Chinese loong boat (ιΎθ), a symbol of coordinated power and forward momentum
Want to see what else we are building? Explore our other open-source projects on the baidu-baige GitHub organization
We'd love to hear from you β reach us through any of these channels
Join the developer group β scan the QR code posted in our GitHub issue
Follow us on RedNote (ε°ηΊ’δΉ¦) for release notes and practice sharing
loongforge@baidu.com β for collaboration, adoption, and any other enquiries
Chat with the team and other users in our Slack workspace
LoongForge is released under the Apache License 2.0. Some files are derived from third-party open-source projects β please consult file headers for their specific notices
@software{LoongForge2026,
title = {LoongForge: A modular, scalable, high-performance training framework for LLMs, VLMs, diffusion, and embodied models},
author = {{The LoongForge Authors}},
year = {2026},
url = {https://github.com/baidu-baige/LoongForge}
}LoongForge is built upon NVIDIA's Megatron-LM. We also referenced and drew inspiration from excellent open-source projects including Transformers, LLaMA-Factory, and Megatron-Bridge. We sincerely thank these communities for their outstanding contributions