Reflection AI launches Beam, its first open-source model, to compete with Chinese research laboratories

The New York-based start-up founded by former DeepMind employees claims to match GLM-5.2 and come close to Qwen 3.8-Max in terms of code and agent-based tasks, whilst using three to four times less computational power.

Reflection AI, the company founded in 2024 by former DeepMind researchers Misha Laskin and Ioannis Antonoglou, is releasing its first open-weight model, called Beam. Its CEO presents it as a "workhorse": a model that is particularly solid in programming and agentic tasks, which would need three to four times less computing power than its open competitors to reason about a problem, and therefore faster and cheaper to exploit.

According to the company, Beam is at the level of GLM-5.2, released in June by the Chinese company Z.ai, and is approaching Alibaba's Qwen 3.8-Max, which is still behind the best American closed models. Reflection is aimed at companies and administrations that want to master their own AI systems without resorting to Chinese models. A much more powerful successor is already in training.

A voice for open source in Washington

Backed by Nvidia, Sequoia and Citigroup, and valued at $25 billion before its latest round of funding in June, the start-up joins Mistral and Thinking Machines Lab among the Western players in open models. It says it has Beam evaluated by the American Center for Super-Intelligence Standards and the British AI Safety Institute, and wants to weigh in on the regulatory debate underway in the United States.