Reflection AI unveils Beam open weight model to compete with Chinese labs

TechCrunch AI and MarkTechPost reported that Reflection AI unveiled Beam, its first open-weight artificial intelligence model, on October 5, 2026. The two-year-old startup based in Brooklyn designed the model specifically for coding and agentic workloads, according to TechCrunch AI and MarkTechPost. [1][2]

MarkTechPost reported that Beam is structured as a 501-billion parameter sparse Mixture-of-Experts model with 23 billion active parameters. Reflection AI claimed that Beam matches the reasoning performance of leading Chinese open models like GLM-5.2 while using three to four times less inference compute, according to TechCrunch AI and MarkTechPost. [1][2]

TechCrunch AI reported that Reflection AI is aiming Beam and future models at enterprises and sovereign nations to build AI factories. This pitch allows institutions to construct customized local AI systems by training the models of Reflection AI on their own proprietary data. TechCrunch AI added that the startup aims to provide a Western answer to models such as DeepSeek, Qwen, and Z.ai. [1]

According to MarkTechPost, Reflection AI plans to release the Apache 2.0 weights later in October 2026. GLM is a series of open-weight large language models developed by the Chinese software company Z.ai for tasks including software development. [2][10]