Kimi K3 tests US confidence in AI lead

Kimi K3, a new AI program from Chinese startup Moonshot AI, jolted investors and policy watchers after its Thursday release. Within hours, it reached the top spot on Arena, a closely watched ranking of AI coding tools, making it the first Chinese model to take the number one position on the list.

The launch added pressure to US companies such as OpenAI and Anthropic, whose paid, closed-source services face a challenge from lower-cost Chinese models that programmers can customize. The reaction echoed early 2025, when DeepSeek unsettled markets by releasing a powerful model at a fraction of the usual cost, briefly wiping hundreds of billions of dollars off the value of US tech companies.

Anastasios Angelopoulos, who runs Arena, said Kimi K3 could “cause a reckoning in the capital markets” as businesses weigh free, locally customizable Chinese systems against American models that require sharing data with outside companies. Venture capitalist and White House AI adviser David Sacks argued on X that limits on data centers and model releases risk weakening the US in the AI race.

Dean Ball, a former White House AI adviser now at OpenAI, said Kimi K3 appeared to be a strong program rather than a copy of American models. He predicted the Trump administration may discourage US companies from using Chinese AI by warning of hidden risks, while investor Gavin Baker described the release as potentially negative for OpenAI and Anthropic but broadly positive for other companies.