For several weeks this summer, the AI ​​industry was obsessed with Anthropic’s latest frontier models and Washington’s fight to control who was granted access to them. But while everyone was watching the frontier, developers kept building, and they weren’t waiting for permission from the Anthropics and OpenAIs of the world.

Chinese open-weight models accounted for 41% of downloads on Hugging Face this spring, surpassing American models. On OpenRouter, the six most popular models are all open models from Chinese companies, including Tencent, Xiaomi, DeepSeek, MiniMax, and Z.ai. Anthropic’s Claude Opus 4.7 is in seventh place, as of this writing. And Vercel data shows that open models are absorbing much of the high-volume infrastructure of AI applications, while closed models operate as the highest-cost premium layer. Open models handled nearly a third of AI requests on the platform in June.

Those platforms only capture a portion of the AI ​​ecosystem; in particular, they leave out sessions hosted by major labs, which likely account for the majority of OpenAI and Anthropic usage. But the large and growing market share of open source models raises a difficult question: How much do frontier models still matter if most production AI ends up running on cheaper, customizable alternatives?

Some see the growth of open source models as a sign that smarter models may end up being used only for the most specialized use cases. “Maybe in a few years, frontier models will be for experimenting and [for] some very high-value tasks, and most production workloads will actually be driven by proprietary models within companies or by open source models,” Clem Delangue, CEO of Hugging Face, said on a recent episode of Equity.

Hugging Face is a platform and developer community best known for hosting, sharing, and helping businesses implement open models. Delangue says Hugging Face customers and community members are increasingly touting the benefits of owning their own AI models rather than renting them, a trend that has gained traction in the cold light of day after receiving the bill associated with the cost of scaling closed-frontier models.

“If you’re an AI company or a technology company, you don’t want to outsource your core capabilities to another company, to a black box API that you don’t control, that you don’t have any visibility over and that you don’t really have any ownership,” Delangue said.

That change, Delangue maintains, is reflected in the activity that occurs at Hugging Face. A new repository is created every seven seconds on the platform, hosting nearly three million public models and one million public datasets, according to Delangue. That points to a different image than “one model that rules them all,” he says. In reality, it seems more like companies use many different models, many of which are customized for their specific use case. Half of all Fortune 500 companies are using Hugging Face to implement their own proprietary models and open source models, he says.

The growing popularity of open models coincides with a steady stream of increasingly capable releases from Chinese AI labs.

Every few months, another Chinese AI company releases a powerful open model that is cheaper to deploy and easier to customize than closed competitors, undermining the proprietary AI economics that American companies have invested billions in. More recently, Beijing-based artificial intelligence company Z.ai released an open model called GLM-5.2 that excels at agent coding and competes with Anthropic’s latest models in identifying security vulnerabilities.

Delangue is not the only executive who argues that companies should avoid being tied to a single supplier model.

Microsoft CEO Satya Nadella recently warned against locking in to a single vendor, arguing that data control should be a top concern for companies using AI.

“While the great innovation that comes from model vendors having fair use rights to train models with public data is needed, I find it ironic that the status quo is to then turn around and impose restrictive terms on distillation, and reserve the right to learn from customer interaction and usage data,” Nadella said. “If learning flows in only one direction, the economic value converges towards the owners of the learning infrastructure rather than the creators of the knowledge itself. Therefore, it is imperative that we distribute the learning infrastructure to each company so that they can control their own learning cycle.”

The rise of open models has also intensified debate over whether increasingly capable models should be widely available.

Anthropic CEO Dario Amodei has argued that scaling weights of powerful open models could become dangerous because once they are launched, they become difficult to control. Others have argued that bad actors can more easily access open models and could use them to spread disinformation or implement cyber or biological warfare.

Delangue sees compensation differently.

“The biggest risk in AI is the concentration of power,” Delangue said. “In my opinion, the way to make the world safer is to level the playing field and create transparency in these models.”

Transparency means defenders can more easily “patch cybersecurity risks that they already know open source models can exploit,” he said.

The Hugging Face executive maintains that keeping powerful models closed does not eliminate the risks associated with advanced AI systems, in part because it is easy to bypass the API security barriers of frontier models and steal the weights and spread them openly. Restricting powerful models, Delangue argues, simply concentrates technology in the hands of a few companies while reducing transparency about how the systems work.

“It really doesn’t make it safe to keep it behind closed doors for just a few players,” Delangue said. “You make it more dangerous because you create asymmetry of power and asymmetry of capabilities.”