This summer, the AI world was glued to Anthropic’s latest frontier models and the Washington tug-of-war over who could use them. But behind the headlines, developers kept shipping — and they weren’t waiting for permission from big labs. Chinese open-weight models grabbed 41% of downloads on Hugging Face this spring, overtaking U.S. offerings. On OpenRouter, the top six most popular models are all open releases from Chinese firms like Tencent, Xiaomi, and DeepSeek. Anthropic’s Claude Opus 4.7 sits at seventh. Vercel data shows open models now handle nearly a third of AI requests on its platform, acting as the workhorse layer while closed models serve as a premium, high-cost tier.
Those numbers don’t capture everything — they miss the internal sessions at major labs that likely make up most of OpenAI and Anthropic’s usage. But the trend raises an uncomfortable question: How much do frontier models really matter if most production AI runs on cheaper, customizable alternatives? Hugging Face CEO Clem Delangue thinks the answer is not much. “Maybe in a few years, frontier models will be for experimenting and high-value tasks, while most production workloads run on private or open source models,” he said. Delangue sees companies moving away from renting black-box APIs toward owning their own models, especially after seeing the bills for scaling closed systems. A new repository is created every seven seconds on Hugging Face, which now hosts nearly three million public models. Half of all Fortune 500 firms use the platform to deploy private or open models.
The open-model surge is being fueled by a steady drumbeat of capable releases from Chinese labs, each undercutting proprietary AI’s economics. Beijing-based Z.ai recently dropped GLM-5.2, an open-weight model that competes with Anthropic’s latest on security vulnerability detection. Microsoft CEO Satya Nadella has also warned against single-provider lock-in, arguing firms need to control their own learning loops. Meanwhile, the debate over open models’ safety rages. Anthropic’s Dario Amodei warns they could enable bad actors; Delangue counters that the real danger is concentrating power. “You don’t make it safe by keeping it behind closed doors,” he said. “You make it more dangerous by creating asymmetry of power.”
Source: TechCrunch