Stanford AI Index 2026: US-China model gap down to 2.7%

Industry produced more than 90% of notable frontier models released in 2025, and by March 2026 the top US model led its nearest Chinese rival by just 2.7%, according to Stanford HAI's 2026 AI Index. DeepSeek-R1 matched the top US model in February 2025, and the report finds six labs (Anthropic, xAI, Google, OpenAI, Alibaba, and DeepSeek) now clustered within 25 Elo points of each other on the Arena Leaderboard, shifting competition toward cost and reliability rather than raw capability.

The result is what the report calls a "jagged frontier": Google's Gemini Deep Think won gold at the International Mathematical Olympiad, while the top model still reads an analog clock correctly only 50.1% of the time. AI agents rose from 12% to roughly 66% task success on OSWorld, and SWE-bench Verified performance jumped from 60% to near 100% in a single year.

"The data does not point in a single direction," write co-chairs Yolanda Gil and Raymond Perrault. "It reveals a field that is scaling faster than the systems around it can adapt."

The money and infrastructure picture is lopsided the other way. US private AI investment reached $285.9 billion in 2025, roughly 23 times China's $12.4 billion, and the US hosts 5,427 data centers, more than 10 times any other country. Global corporate AI investment hit $581.7 billion, up 130% year over year. Yet AI researchers relocating to the US have fallen 89% since 2017, with 80% of that drop coming in the last year alone.

Documented AI incidents climbed to 362 from 233 a year earlier. Organizational adoption reached 88%, and generative AI hit 53% population adoption within three years, faster than the PC or the internet. Public opinion still trails expert opinion: 73% of AI experts expect a positive job impact against 23% of the general public, a 50-point gap.