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

Open-source AI models now rival proprietary ones in key areas while costing less to run.

Reflection AI, a startup founded by former DeepMind researchers, has launched its first model, Beam. This open-source model is positioned as a Western alternative to leading Chinese AI models. Beam reportedly rivals GLM 5.2 in reasoning capabilities while consuming significantly less compute power, using three to four times less. It also approaches the performance of Alibaba's Qwen 3.8-Max on coding and agentic tasks. The company aims to serve businesses and governments seeking to develop sovereign AI systems without relying on Chinese models, addressing security concerns associated with open-source AI that can be downloaded and modified with fewer safeguards. Reflection AI has received investment from prominent firms including Nvidia, Sequoia, and Citigroup, and is already developing a more powerful successor to Beam.

Beam's performance benchmarks, particularly its efficiency in reasoning and its strength in coding and agentic tasks, suggest a significant leap in open-source model capabilities. This development challenges the perceived superiority of closed, proprietary models and indicates that Western startups are closing the gap with Chinese AI leaders in the open-source domain. The model's lower compute requirements could democratize access to advanced AI for smaller organizations and researchers.

The launch of Beam, coupled with Reflection AI's focus on sovereign AI solutions, points to a growing market demand for controllable and secure AI systems outside of Chinese-developed models. This positions Reflection AI and similar Western startups as key players in shaping the future of AI regulation and deployment, particularly as governments grapple with AI safety and geopolitical implications. The company's engagement with US and UK AI safety institutes underscores the increasing importance of collaborative efforts in developing responsible AI governance.