An open research initiative
for the reinforcement learning era.
RL Commons brings together researchers, compute partners, and contributors around the problems slowing reinforcement learning down — compute access, shared environments, and visibility for good work.
The RL research community is facing real problems right now.
AI research is entering a phase in which post-training, reinforcement learning, reasoning, and agent systems matter more than ever. Many of the most important breakthroughs over the next few years will not come from the largest pretraining runs. They will come from researchers with sharp ideas about environments, rewards, evaluation, and training dynamics.
But in practice, three specific problems are holding the field back:
Compute access. Most labs can't run the experiments they want to. R1-style post-training is out of reach for most university teams — not because the ideas aren't there, but because the resources aren't.
Environments and benchmarks. RL progress depends on shared evaluation ground. Today, too much of that work gets rebuilt privately in every lab, and the best contributions don't travel.
Visibility. Strong research gets done, then gets buried. There's no good connective tissue for the community to find, amplify, and build on each other's work.
RL Commons is our attempt to address these together. Modeled on the organizing principles of academic workshops, we bring together researchers, compute partners, and infrastructure builders around a shared mandate: pool resources, lower barriers, and make serious RL research easier to actually run.
Three things we're building.
Compute for
research teams
A research platform connecting researchers with compute providers, so teams can actually run the experiments they've been putting off. First delivered through cohort programs like Project Aster.
Environments
& benchmarks
An open library of RL environments, benchmarks, and rubrics — where contributors get credit, visibility, and a real audience for their work. Shared reference points the field can build on.
A community
for RL research
A connective tissue for RL research — helping good work get seen, shared, and built upon across university labs, open-source contributors, and the broader community.
A reciprocal effort.
Free compute. Your research. Real audience.
Free managed RL infrastructure through cohort programs, a cohort of peers, full ownership of your work, and active promotion of your research across our channels. Project details remain confidential by default.
Direct access to the next generation of RL.
Direct connection to the researchers shaping the next five years of RL, founding-partner visibility across every program output, and the opportunity to support open science at a formative moment for the field.
Credit, attribution, and a real audience.
For the researchers and engineers who contribute environments, benchmarks, and tools to the open library: durable credit, citation-friendly attribution, and direct amplification to the community that needs your work.
Project Aster — the founding cohort.
Project Aster
A nine-week founding cohort providing five to ten selected research teams with free access to managed RL infrastructure. Teams focus on the research — environments, rewards, training strategy, evaluation — while the infrastructure partner handles distributed execution, rollout orchestration, and scheduling.
Scope: RL-only projects on 1.5B–3B open-weight models. Reasoning, reward design, environment design, agentic learning, policy optimization, and evaluation. Teams retain full ownership of their work; project details are confidential by default.
The founding circle is open.
RL Commons is in its founding phase. We are actively inviting compute partners, research institutions, and advisors to join as founding members. The circle is intentionally small at launch — we prioritize quality of participation over breadth.
Logits
Provides the distributed RL execution engine (Echo2) and the researcher-facing platform behind Project Aster. Initiated RL Commons and contributes ongoing operational support.
Invitation open
For GPU providers committing capacity to the first cohort. Founding partners are recognized across program materials and shape the evolution of the compute pool.
Invitation open
For research institutions, faculty groups, and open-source labs joining the founding cohort. Shape the research agenda from the first cycle.
There are several ways in.
Frequently asked, honestly answered.
Is RL Commons a non-profit?
Not formally. RL Commons is an open research initiative — a community-oriented collective of researchers, compute partners, and infrastructure partners. Whether to formalize as a non-profit is something we may revisit as the initiative matures.
Do researchers retain ownership of their work?
Yes. Teams retain full ownership. Project details, unpublished results, and experimental progress are treated as confidential by default and are not shared across teams, with compute partners, or externally unless the team explicitly consents.
Who is eligible for Project Aster?
University labs, faculty-led research groups, student research organizations, independent open-source researchers, and small teams with a serious RL research question. The strongest fit is usually not the biggest lab — it is the team with a clear question and a real plan to use the cycle well.
Can our lab / company become a partner?
Yes. We are actively inviting founding compute partners and research institutions. If you are exploring a partnership, please reach out.