Beam (by Reflection)

Beam is Reflection's first open-weight model, a sparse Mixture-of-Experts LLM with 501B total and 23B active parameters, focused on coding, reasoning, and agentic workloads. It is not yet generally available; weights, model card, and a technical report are promised later this month, with early API access through Reflection's platform. The announcement emphasizes two differentiators: competitive benchmark results against large open models such as GLM, Qwen, Kimi, and DeepSeek, and inference efficiency from low active parameters. Revenue, pricing, and growth are not disclosed. For founders, Beam is less a SaaS business than an infrastructure provider play: its success depends on developer adoption of the API, enterprise trust, and whether open weights undermine paid API demand.