UC Berkeley AgentBeats competition backs Hugging Face agentic RL challenge
University of California, Berkeley’s AgentBeats competition is hosting a new reinforcement learning track, announced on LinkedIn by Hugging Face, with $10,000 in credits up for grabs.
Hugging Face has taken to LinkedIn to announce a new agentic reinforcement learning challenge in partnership with PyTorch and Unsloth, offering a $10,000 prize pool in Hugging Face credits.
Hugging Face builds open source tools and platforms for developing, training, and sharing machine learning models, with a growing focus on education, community-led learning, and developer tooling. The new OpenEnv Challenge sits within the University of California, Berkeley AgentBeats Competition and introduces a dedicated reinforcement learning track.
Ben Burtenshaw, community education in AI at Hugging Face, shared the announcement on LinkedIn, writing, “[Agentic RL Hackathon] We're partnering with @PyTorch and @UnslothAI on the OpenEnv Challenge! Build Agentic RL to win $10K in HF credits!”
Focus on environments, not models
Unlike many AI competitions that emphasize model performance, the OpenEnv Challenge centers on the design of reinforcement learning environments.
Participants are asked to create an RL environment and publish it to the Hugging Face Hub, share training code, and write a technical blog outlining their approach. Selected entries will also have the opportunity to publish on the PyTorch blog.
Burtenshaw noted that the challenge includes “a special track just for reinforcement learning,” positioning environment design as core infrastructure for agent-based systems.
The challenge is built around OpenEnv, a framework designed to standardize how agentic reinforcement learning environments are created and deployed. OpenEnv uses Gymnasium-style APIs and is intended to support interoperability across modern PyTorch-native tooling.
According to the challenge description, submissions will be evaluated on creativity, technical quality, scalability, and alignment with OpenEnv’s specification and hub ecosystem. Hugging Face says winning projects will be open source and contribute directly to shared benchmarks and learning resources.
Burtenshaw emphasized the collaborative nature of the initiative, writing, “Additionally, we’re thrilled to announce a new AgentBeats custom track: the OpenEnv Challenge: SOTA Environments to Drive General Intelligence.”
Credits, publication, and community exposure
The prize pool totals $10,000 in Hugging Face credits, sponsored by Hugging Face, the PyTorch team at Meta, and Unsloth. Beyond credits, participants can gain visibility through publication opportunities and integration into the OpenEnv hub.
The competition is positioned as a hands-on learning opportunity for students and early-career practitioners working in agentic AI and reinforcement learning.
Burtenshaw framed the motivation behind the initiative on LinkedIn, writing, “Build Agentic RL to win $10K in HF credits!”
ETIH Innovation Awards 2026
The ETIH Innovation Awards 2026 are now open and recognize education technology organizations delivering measurable impact across K–12, higher education, and lifelong learning. The awards are open to entries from the UK, the Americas, and internationally, with submissions assessed on evidence of outcomes and real-world application.
