Reflection AI Introduces Beam, a 501B Open-Weight Model Trained to Spend Fewer Tokens

Reflection AI has introduced Beam, its first open-weight model, pitching it as a token-efficient system for coding, reasoning and agentic workloads. The company says Beam is a sparse Mixture-of-Experts model with 501 billion total parameters and 23 billion active parameters, trained so successful answers are rewarded not only for correctness but also for avoiding unnecessary tokens.