Beam Beam is Reflection AI’s first open-weight model: a text-only sparse mixture-of-experts system with 501 billion total parameters and 23 billion active, a 1 million token context window, and a focus on coding, reasoning, and agentic workloads. It was pretrained on 23.8 trillion tokens and post-trained with large-scale reinforcement learning. Early access is on a waitlist, with Apache 2.0 weights planned later in October 2026.