A training framework born from real-world, large-scale production workloads โ and shared back with the community
LoongForge didn't start as an open-source project. It grew out of AIAK-Training โ Baidu Baige's commercial training acceleration offering delivered to enterprise customers (previously closed-source) โ after years of hardening under real production workloads
Before going open source, LoongForge was already powering large-scale models in production:
It now joins the Baige Loong open-source series โ named after the traditional Chinese loong boat (้พ่), a symbol of coordinated power and forward momentum. Sister project: LoongFlow โ A Thinking & Learning Framework for Expert-Grade AI Agents
A glimpse of what is next
LoongForge is released under the Apache License 2.0. Some files are derived from third-party open-source projects โ 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