Is China quietly winning the artificial intelligence race?

 

AI-generated 

Is China quietly winning the artificial intelligence race?

Matt Madrigal, Pinterest's chief technology officer, said the strength of these models lies in the fact that they can be freely downloaded and customized by companies like his, which is not the case for the majority of models offered by American competitors such as OpenAI, the developer of the ChatGBT chatbot.

Madrigal added: "The open-source technologies we use to train our internal models are 30% more accurate than leading off-the-shelf models."
He explained that these improved recommendations come at a much lower cost, sometimes up to 90% less than using the closed models preferred by American AI developers.

"Fast and cheap"

Pinterest is not the only American company relying on Chinese artificial intelligence technologies; these models are gaining momentum among a number of Fortune 500 companies.

Airbnb CEO Brian Chesky told Bloomberg in October that his company had relied "a lot" on Alibaba's Coin model to power its smart customer service agent.
Chesky offered three simple reasons for this: the model is "excellent," "fast," and "cheap."

Further evidence of this can be found on the "Hugging Face" platform, which users turn to for downloading ready-made AI models, including models from major development companies such as Meta and Alibaba.

Jeff Bodie, who develops products for the platform, said that the cost factor is what drives startups to prefer Chinese models over their American counterparts.
He added: "If you look at the most popular models on Hugging Face - those with the most downloads and likes - Chinese models from Chinese labs are usually in the top ten."

He continued: "There are weeks when four out of five leading training models on Hugging Face are from Chinese labs."
In September, Coin's model surpassed Meta's Llama model to become the most downloaded big language model on the Hugging Face platform.

Meta launched its open-source Llama AI models in 2023. Until the release of DeepSec and Alibaba's models, they were considered the primary choice for developers working on custom applications.

But the launch of Llama 4 models last year disappointed developers, and according to reports, Meta has begun using open-source models with Alibaba, Google and OpenAI to train a new set of models scheduled for release this spring.

Chinese success

As 2025 began, the prevailing view was that despite American technology companies spending billions of dollars, Chinese companies threatened to overtake them.
"That's no longer the story," said Bodie, adding, "The best model now is an open-source model."

A report published by Stanford University last month concluded that Chinese artificial intelligence models "appear to have caught up with or even surpassed" their global counterparts, both in terms of their capabilities and the number of their users.
In a recent interview with the BBC, former British Deputy Prime Minister Sir Nick Clegg said he felt that American companies were overly focused on pursuing artificial intelligence that might one day surpass human intelligence.

Some experts now describe these ambitions as vague and ill-defined, giving China an opportunity to dominate the field of open-source artificial intelligence.
The Stanford University report also indicated that China’s success in developing open-source models may be partly due to government support.

On the other side of the world, American companies like OpenAI are under enormous pressure to increase their revenue and achieve profitability, and are now turning to advertising to do so.