2 Chinese AI models gain US foothold as Mozilla CTO, Coinbase switch

For AI developers, the draw of Moonshot’s Kimi K3 isn’t just its lower price tag—it’s the raw performance that Mozilla’s CTO says ‘just seems snappier’ than Anthropic’s polished Claude Fable. Combined with Z.ai’s GLM-5.2, these models are convincing pragmatic engineers to overlook geopolitical risks, reshaping the competitive landscape for frontier AI.

Chinese AI models are rapidly gaining traction in the United States, marking a significant shift in the global artificial intelligence landscape. Just over a year after DeepSeek’s debut rattled the industry, a new wave of advanced, low-cost models from Chinese startups is attracting high-profile American users—from Mozilla’s chief technology officer to cryptocurrency exchange Coinbase. Raffi Krikorian, CTO at Mozilla, publicly disclosed that he switched to Moonshot’s new Kimi K3 model for many daily tasks within days of its launch, praising its responsiveness compared to Anthropic’s premium Claude Fable. He had already been using Z.ai’s GLM-5.2 for routine calendar, document and email management. These endorsements highlight a pragmatic shift: U.S. developers and enterprises are prioritizing cost efficiency and performance, even as geopolitical tensions simmer.

“

Raffi Krikorian, CTO at Mozilla, publicly disclosed that he switched to Moonshot’s new Kimi K3 model for many daily tasks within days of its launch, praising its responsiveness compared to Anthropic’s premium Claude Fable.

The appeal is straightforward. Chinese models are dramatically cheaper than many U.S. counterparts, offering capabilities that in some cases match or exceed pricier alternatives. Krikorian described K3 as “snappier,” a qualitative metric but one that matters for real-time applications. Coinbase confirmed it is transitioning to Chinese AI to cut operational costs, joining a growing cohort of Western firms exploring these models. The open-access nature of some Chinese releases also allows extensive customization, a key advantage for independent developers. This economic calculus is reshaping the competitive dynamics: if U.S. companies cannot deliver comparable value, the market will increasingly look east.

Yet the rise of Chinese AI is embroiled in controversy. The Trump administration, on July 22, 2026, accused Moonshot of using “covert” methods to build K3 by leveraging Anthropic’s technology—specifically, illicit distillation from Claude Fable. Anthropic and some U.S. lawmakers have repeatedly raised allegations of intellectual property theft, claims Beijing dismisses as “groundless.” Treasury Secretary Scott Bessent has signaled more sanctions could target Chinese AI to protect American IP. These political headwinds create a precarious environment: while the technology is technically accessible now, future bans or export controls could abruptly cut off access, leaving adopters stranded.

This tension underscores a fundamental dilemma facing U.S. policymakers. The 2022 export restrictions on advanced chips aimed to slow China’s progress, but the domestic breakthroughs in model efficiency suggest those curbs may have inadvertently spurred innovation in software optimization. Chinese startups have learned to do more with less, producing models that are both powerful and economical. The current situation mirrors the broader decoupling debate: can the U.S. maintain its lead through sanctions, or will it simply cede a growing global market to Chinese alternatives? For now, the absence of an outright ban means the models remain available, and their adoption is spreading not just among hobbyists but within critical infrastructure firms like a major crypto exchange.