Reflection Launches Beam, Defying Chinese Giants with Open-Weight Efficiency
Reflection has released Beam, a 501-billion parameter mixture-of-experts model that activates just 23 billion parameters per token. Designed to rival top Chinese open-weight systems like DeepSeek and Qwen, Beam prioritizes high efficiency, aiming to match advanced coding and reasoning benchmarks while consuming a fraction of the computational power.
Source: The Decoder