The U.S. has a new horse in the race to build the best open-source AI model

Tuesday, October 6, 2026

Andrew Nusca

Editorial Director, Brainstorm

Profile

Good morning. If there’s anyone who understands a war of attrition, it’s the U.S. military.

The Pentagon has reportedly halted using Anthropic’s Claude, finally, after blacklisting the San Francisco AI company in February and triumphing over a lawsuit about it in September. The delay is believed to be at least partly because of Claude’s deep integration in a Palantir platform that the Pentagon continues to rely on.

Though it’s unfair to say that the two sides have been exchanging blows all year—it’s a rather one-sided aggression from the feds—it is fair to say that Anthropic’s conflict with the Department of Defense adds up to a sort of litmus test: Are the company’s AI wares distinctive enough (or embedded enough) to, in their absence, substantially harm the intelligence gathering operations of one of the largest employers in the U.S.?

It’s a good question as Anthropic heads toward its long-awaited IPO. You need not prompt your favorite generative AI model to know that investors won’t support a company’s $2 trillion valuation if its products are rapidly becoming low-margin commodities.

But hey, maybe I’m just hallucinating. Today’s tech news follows. —Andrew Nusca

Thoughts? Suggestions? Hot tip? Drop me a line.

Were you forwarded Fortune Tech? Sign up here.

THE BIG STORY

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

Reflection AI CEO Misha Laskin in San Jose, California on March 18, 2026. (Photo: David Paul Morris/Bloomberg/Getty Images)David Paul Morris/Bloomberg/Getty Images

The U.S. has a new horse in the race to build the best open-source AI model—a competition China is currently dominating with models such as Moonshot’s Kimi K3 and DeepSeek’s V4.

The AI company Reflection, on Monday, unveiled its first model, called Beam. It “advances the frontier for the Western open ecosystem,” the New York company told Fortune.

The company said that in benchmark tests it conducted, Beam achieved scores similar to those of GLM-5.2, an open model from Beijing-based Z.ai that was released back in July. That model was succeeded by GLM-5.3, which Z.ai debuted in August.

Reflection said Beam is cheaper to run than many competing models.

“Through high-compute reinforcement learning, we were able to make Beam extremely efficient at reasoning—delivering competitive performance on coding and agentic tasks at a fraction of the token cost and inference time compute,” the company said.

Reflection added that Beam is three to four times as efficient as rival open models from Western companies.

It is positioning Beam as a “workhorse” for enterprises, the public sector, and developers. In May, the company entered into a partnership to supply its models to the U.S. Department of Energy and the U.S. Department of War.

Reflection was founded in March 2024 by former Google DeepMind researchers Misha Laskin, now its CEO, and Ioannis Antonoglou, the startup’s president and chief technology officer. The Nvidia-backed firm has a $25 billion valuation, Laskin told CNBC in April.

The Western company most comparable to Reflection may be Thinking Machines Lab, founded by OpenAI’s former CTO Mira Murati. It unveiled its first model, Inkling, in July. Nvidia also makes an open model, called Nemotron, and OpenAI, Google, and Meta all offer open models as well. —Emily Forlini