Reflection unveils 501B-parameter open-weight Beam model, pledges Apache 2.0 release this month

Reflection unveils Beam, a 501-billion-parameter Mixture-of-Experts reasoning model with 23 billion active parameters, and says it will release Apache 2.0-licensed weights and developer tools later this month.

• Reflection introduces Beam, a 501-billion-parameter open-weight Mixture-of-Experts model with 23 billion active parameters, and plans to release its Apache 2.0-licensed weights and developer tools later this month.
• Beam’s reinforcement-learning campaign generates more than 100 million rollouts over four weeks on 10,500 NVIDIA GB300 GPUs, using roughly 1.3 billion sandbox runs.
• Reflection says Beam matches GLM-5.2-level advanced reasoning while using 3–4 times less inference compute, and supports a reasoning-effort setting that trades response length for performance.