Reflection debuts Beam, a open-weight AI model to rival Chinese models at lower compute cost

On Monday, October 5, 2026, Brooklyn startup Reflection AI announced the launch of Beam, its first frontier, open-weight artificial intelligence model. Founded in 2024 by former Google DeepMind researchers, the two-year-old company has raised roughly $4.7 billion from backers like Nvidia, Sequoia Capital, and Lightspeed Venture Partners, reaching a $25 billion pre-money valuation.

Beam is a text-only, mixture-of-experts model featuring 501 billion total parameters, 23 billion active parameters, a 1 million token context window, and pre-training on 23.8 trillion tokens. According to Reflection, Beam matches the reasoning performance of leading Chinese open models like Z.ai’s GLM-5.2 while using significantly less inference compute. The company also states that Beam outperforms top Western open models and outscores Mira Murati’s Thinking Machines Lab model, Inkling, on shared coding tests.

Beam is aimed at businesses, public agencies, and developers to handle reasoning, coding, and agent tasks. Reflection plans to release Beam’s model weights and technical details later this month, with availability through major cloud providers and open-source libraries.