Reflection AI debuts Beam to challenge Chinese open models at lower compute cost

Brooklyn-based Reflection AI on Monday formally unveiled Beam, its first frontier open-weight artificial intelligence model. The two-year-old startup says Beam matches leading Chinese open models on advanced reasoning benchmarks while running at a fraction of their token and inference-compute cost — a claim that sharpens the Western race to answer DeepSeek, Qwen and Z.ai.

According to TechCrunch, Beam is a text-only mixture-of-experts system with 501 billion total parameters and 23 billion active. It was pretrained on 23.8 trillion tokens and offers a 1 million-token context window. Reflection positions Beam as a “workhorse” for enterprises, the public sector and developers that need coding, reasoning and agentic performance without closed-lab lock-in.

The company says Beam scores on par with Z.ai’s GLM-5.2 and outperforms leading Western open models while using 3–4× less inference compute. Those figures have not been independently verified. Reflection was founded in 2024 by two former Google DeepMind researchers and has raised roughly $4.7 billion from backers including Nvidia, Sequoia Capital and Lightspeed Venture Partners, with a last-round pre-money valuation near $25 billion.

Reflection plans to release Beam’s weights and full technical details this month through hyperscalers, neoclouds and open-source integrations. It is also pitching “AI factories” that let institutions train customized local systems on proprietary data — including early sovereign-factory testing with South Korea’s Shinsegae Group — as the United States seeks competitive open alternatives to Chinese labs.

Source: TechCrunch