Reflection AI Introduces Beam: A 501B Open-Weight MoE Model With 23B Active Parameters for Coding and Agentic Workloads

Reflection AI has released Beam, a 501B sparse Mixture-of-Experts model designed specifically for coding tasks and agentic workflows. By utilizing 23B active parameters, the model aims to achieve reasoning performance comparable to GLM-5.2 while requiring significantly less inference compute. The company intends to release the weights under an Apache 2.0 license later in October.