NY-based Reflection unveils Beam, an open model it says rivals GLM 5.2 on reasoning, while using 3x-4x less compute, and Qwen3.8-Max on coding and agentic tasks

Western businesses can now deploy advanced open-source AI for reasoning and coding with lower compute costs.

New York-based startup Reflection AI has launched Beam, an open-source AI model designed as a Western alternative to leading Chinese models. Founded by former DeepMind researchers, Reflection AI claims Beam excels in reasoning tasks, rivaling GLM-5.2 with significantly less computing power (3-4x less). The model also shows strong performance in coding and agentic tasks, approaching the capabilities of Alibaba's Qwen 3.8-Max, though still trailing top closed models from OpenAI and Anthropic. Beam aims to serve businesses and governments seeking to build sovereign AI systems without relying on Chinese technology, addressing concerns over security and control associated with open-source models. Reflection AI has secured significant investment from entities like Nvidia and Sequoia, and is already developing a more powerful successor to Beam.

Beam's release intensifies competition in the open-source AI market, directly challenging the dominance of Chinese models which have seen adoption by Western companies for cost savings. Reflection AI's focus on efficiency and performance in coding and agentic tasks positions Beam as a practical "workhorse" for enterprise applications, potentially lowering barriers for organizations hesitant about the resource demands of proprietary models.

The move by Reflection AI, backed by major investors, highlights a growing trend of Western startups developing advanced open-weight models to offer alternatives that address geopolitical and security concerns. As regulatory discussions around AI intensify, the emergence of models like Beam, with their emphasis on transparency and controlled deployment, could influence policy debates and provide businesses with more strategic options for AI adoption.