Models ·
Beam LLM Opens Early Access With Weights Still Pending
Models ·
Beam LLM Opens Early Access With Weights Still Pending
AI summary · Generated from this article
Reflection AI announced selective early access to Beam, a 501-billion-parameter sparse mixture-of-experts model designed for coding and agentic tasks, but the actual model weights remain unavailable. The company claims Beam achieves performance comparable to GLM-5.2 while using three to four times less inference compute, though these claims rely on company-reported benchmarks and an approximate efficiency calculation that excludes prompt processing and serving overhead. Public weights, technical documentation, and deployment artifacts are promised for later in October but have not yet been released.