Last week, Chinese startup Moonshot AI made waves after it launched Kimi K3— a new AI model which it claims is the largest open-weight model in the world and rivals the most advanced models released by American AI titans OpenAI and Anthropic. The model beat OpenAI’s GPT 5.6 Sol and Anthropic’s Fable 5 at coding tasks, according to AI leaderboard Arena. It’s also on par with OpenAI’s top model on agentic tasks, Arena CEO Anastasios Angelopoulos tells me.

“America’s net scientific hegemony has been essentially challenged by this,” Angelopoulos says.

In recent months, Chinese open source models have caught on among American companies, mainly because they’re significantly cheaper and businesses are trying to rein in their AI spending. But these models can also be downloaded, tweaked and controlled to fit a firm’s preferences. And since they run locally, companies don’t need to send their proprietary data to an external company or worry about the model abruptly being banned after building entire systems and applications on top.

“What happened with Fable? Fable was just taken away at a moment's notice. Imagine that you had built your company's workflows on Fable. It's so destructive,” Angelopoulos says.

Add geopolitical tensions to the mix and matters become more urgent. The Trump Administration has discussed banning Chinese open source models in the country while Beijing has also reportedly discussed banning overseas access to its top models.

“I do not believe that American enterprises fundamentally are going to be building on Chinese-made AI models in the long run,” Angelopoulos says. “There's too much risk.”

Angelopoulos notes that American open source companies have fallen dramatically behind. But in recent weeks, a gaggle of American startups have started releasing competitive homegrown alternatives to Chinese open source AI models, Forbes reported. Poolside, which got its start building a powerful coding model, released a new open model today called Laguna and has been silently working on a “model building factory” to churn out models every five weeks. OpenAI former CTO Mira Murati’s Thinking Machine Labs recently launched an open-weight model called........

© Forbes