Reflection debuts Beam, an open-weight AI model to rival Chinese models at lower compute cost
Reflection says Beam can match top Chinese open models while using 3-4x less inference compute.
Beam is a 501-billion-parameter, text-only mixture-of-experts model with 23 billion active parameters and a 1 million token context window. Reflection says it is built for reasoning, coding, and agentic tasks, though its benchmark claims have not been independently verified. The startup plans to release the weights and full technical details this month. Reflection is pitching Beam and future models to enterprises and sovereign nations through customized “AI factory” deployments. TechCrunch AI's note
Beam is a 501-billion-parameter, text-only mixture-of-experts model with 23 billion active parameters and a 1 million token context window. Reflection says it is built for reasoning, coding, and agentic tasks, though its benchmark claims have not been independently verified. The startup plans to release the weights and full technical details this month. Reflection is pitching Beam and future models to enterprises and sovereign nations through customized “AI factory” deployments. TechCrunch AI's note
score 6