Chinese AI models are nearing top US models, but the US still leads

In short: Chinese AI companies are building language models that score close to top US models on many tests, often for much less money, but experts say the US still leads where it matters most.

What's going on

Chinese companies now offer large language models, or LLMs (AI tools that can write and answer questions like a chatbot), that are close in quality to the best American models on many standard tests. The New York Times reports that some Chinese systems are “almost on par” with leading US tools and are sometimes offered at very low cost or even free.

Several outside reports point the same way. Stanford’s 2026 AI Index found the gap between leading US and Chinese models has “nearly erased” on one popular benchmark, with only a small difference in scores by early 2026. Other analyses say Chinese models are especially strong in areas like math, coding, and multilingual tasks.

Cost is a major part of the story. Some Chinese models are priced far below comparable US options, which can make them attractive for companies that want something that is good enough and affordable. Think of it like choosing between two cars that both get you to work, but one costs much less to run.

What to watch

Experts still describe this as “America’s race to lose.” The US keeps an advantage in compute (the raw computing power used to train and run AI), advanced chips, and top AI talent, which helps it build the very best models first. Watch whether cheaper Chinese models spread faster globally, and whether US companies can keep top performance while bringing prices down.

Source: NYTimes