Meta Agents Research Environments (ARE) is a platform for evaluating AI agents in evolving, realistic scenarios. It runs the Gaia2 benchmark containing 800 scenarios across multiple domains, testing agent capabilities in complex multi-step reasoning tasks that require dynamic adaptation as new information becomes available. The platform supports multi-agent systems, LLMs, and reinforcement learning approaches.