Reflection AI launches Beam
On 5 October 2026 Reflection AI introduced Beam, its 501B-parameter open-weight model with 23B active parameters and 1M tokens of effective context. Beam stays in final red-teaming; weights will follow this month.
Why it mattered Beam will ship under Apache 2.0, so anyone can run the 501B-parameter model once Reflection AI releases the weights.
Reflection AI introduced Beam on 5 October 2026. The model totals 501 billion parameters. Only 23 billion parameters stay active. Reflection AI pretrained the model on 23.8 trillion tokens. Midtraining stretched the effective context to 1M tokens. Reinforcement learning ran with a 256K-token maximum context. Reflection AI will release the weights under Apache 2.0. The company still runs final red-teaming and evaluations. It will publish the weights, technical report, model card, and artifacts later this month. A select group can access an early version through the waitlist.