Reflection AI launches beam, a cost-effective open-weight model competing with chinese AI giants

Reflection AI has launched Beam, its first open-weight AI model, claiming it matches the performance of leading Chinese models like Z.ai’s GLM-5.2 while using 3-4x less inference compute. The 501-billion-parameter model, with 23 billion active parameters, was trained on 23.8 trillion tokens and features a 1 million token context window. Reflection positions Beam as a 'workhorse model' for enterprises and developers, emphasizing its efficiency in reasoning, coding, and agentic tasks. The startup, founded in 2024 by former Google DeepMind researchers, has raised $4.7 billion, with a $25 billion pre-money valuation. It has secured $7 billion in compute deals with SpaceX and Nebius for access to Nvidia’s GB300 chips through 2029. Beam’s release this month includes full technical details and distribution through hyperscalers and neoclouds. Reflection targets 'AI factories' for institutions to build customized systems using its models. The company also announced testing a sovereign AI factory partnership with Shinsegae Group in South Korea. While performance claims remain unverified, Beam outperforms Western open models and scores on par with Z.ai’s GLM-5.2 on advanced reasoning benchmarks.