Reflection AI unveils open-weight model Beam to rival Chinese labs
Reflection AI unveiled Beam on Oct 5, a text-only mixture-of-experts model with 501 billion parameters, 23 billion of them active, pretrained on 23.8 trillion tokens with a 1 million token context window. The company says Beam matches Z.ai's GLM-5.2 on reasoning benchmarks using "3-4x less inference compute," and has raised roughly $4.7 billion from backers including Nvidia, Sequoia Capital and Lightspeed Venture Partners, according to PitchBook data reported by TechCrunch.
NoteA Western open-weight model claiming Chinese-rival performance at a fraction of the compute cost could lower the price of running frontier-grade AI in-house, a real alternative for banks and insurers weighing proprietary deployments against API access to closed models.