Nvidia-backed Reflection AI unveils its first open model, Beam. Could it be America’s best chance to compete with China?

American AI startup Reflection AI has introduced its inaugural open-source model, Beam, which experts believe could position the U.S. as a strong contender in the global AI model race, currently dominated by China. The company, headquartered in New York City, claims Beam outperforms GLM-5.2, an open model from Beijing-based Z.ai, in benchmark tests.

This achievement is attributed to high-compute reinforcement learning, making Beam highly efficient in reasoning tasks while significantly reducing token costs and computational time.

Reflection AI boasts a $25 billion valuation, led by former Google DeepMind researchers Misha Laskin as CEO and Ioannis Antonoglou as president and chief technology officer. The company has already secured significant deals, including a $6.3 billion compute arrangement with SpaceX for access to Nvidia GB300 chips at its Colossus 2 data center, and a $1 billion deal with cloud provider Nebius in July. These partnerships underscore Beam's potential for enterprise, public sector, and developer applications.

Open-source AI models, such as Beam, can be fully downloaded, run on personal or cloud hardware, and customized to users' needs. This contrasts with proprietary models typically accessed via APIs, limiting user control over data security and costs. As enterprises seek greater data security and cost control, open models are gaining traction.

The company's mission emphasizes open systems' advantages, including rapid evolution, broader participation, and enhanced safety due to community oversight and vulnerability identification.

While Nvidia's Nemotron model currently leads the US-based open model rankings on OpenRouter's leaderboard, other notable models include DeepSeek, Z.AI, Xiamo, Tencent, and Nvidia. OpenAI’s gpt-oss and Google DeepMind’s models also rank among the top contenders. Looking ahead, Reflection AI plans to develop an even larger model, further solidifying its role in the competitive open-source AI landscape.