To push the open source frontier for RL + LLMs, we need scalable, modular environments with real-world complexity, beyond math benchmarks. Today, we’re releasing *benchmax*. An open-source framework to build, run, & scale useful RL envs for LLM fine-tuning, with integrations to