Reflection debuts Beam, an open-weight AI model to rival Chinese models at lower compute cost
Which summary reads better? Pick one — models revealed after.Both summaries are AI-generated.
The new open-weight model Beam delivers frontier-level reasoning and coding performance across a 1 million token context window using a 501-billion-parameter MoE architecture with only 23 billion active parameters, claiming a three-to-four-fold reduction in inference compute compared to rivals. For production teams running agentic workflows, this enables self-hosting highly complex, long-context reasoning pipelines locally at a fraction of the typical hardware footprint and token cost.
Reflection's Beam model achieves comparable performance to leading Chinese models like Z.ai's GLM-5.2 on advanced reasoning benchmarks while using 3-4 times less inference compute, enabling enterprises and developers to deploy high-performance AI at significantly lower operational costs. This development directly impacts the cost and feasibility of building customized AI systems for enterprises and sovereign nations.