The authoritative AReaL roadmap is maintained in the main project repository at areal-project/AReaL. This file is a pointer so contributors browsing the community repository can find it easily.
- Current roadmap document:
AReaL/ROADMAP.md - Quarterly tracking issues: see the
issues tagged
roadmapin the main repository. - Releases: github.com/areal-project/AReaL/releases.
- Release cadence: minor releases roughly every two weeks, major releases roughly
every quarter. See the main
ROADMAP.mdfor the latest cadence.
We welcome community input. The recommended flow is:
- Search existing issues and discussions for prior work on your idea.
- Open a discussion in the Ideas category to gather feedback.
- File a feature request issue once the idea is concrete.
- Indicate support for existing items with 👍 reactions.
For larger contributions, please coordinate with maintainers before submitting a pull request — see CONTRIBUTING.md and the main project's contributing guide.
Our vision for AReaL is to be the go-to framework for training reasoning and agentic AI systems that is:
- Accessible — easy to get started for both researchers and practitioners.
- Scalable — runs from a single workstation to multi-thousand-GPU clusters.
- Flexible — supports diverse algorithms, models, and agentic workflows.
- Performant — competitive training speed and efficiency.
- Open — fully open-source with transparent development.