Stanford 2026 AI Index

AI performance continues to accelerate.

• Industry produced over 90% of notable frontier models in 2025, with some models surpassing human baselines on PhD-level science problems, multimodal reasoning, and competition math.
• SWE-bench Verified performance jumped from 60% to nearly 100%, organizational adoption reached 88%, and 4 out of 5 university students use generative AI.
• The performance gap between US and Chinese models has effectively disappeared, with the leading gap at just 2.7% as of March 2026. The US maintains an edge in 5,427 data centers and investment, but the inflow of researchers and developers has fallen 89% compared to 2017.
• AI agents rose to 66.3% on OSWorld but still fall short of the human baseline of 72.35%, and accuracy on reading analog clocks reached only 50.1%.
• Responsible AI metrics lagged behind. Documented AI incidents rose to 362 cases, and generative AI reached 53% of the world's population in just three years. Annual value for US consumers is estimated at $172 billion, and job-impact optimism stands at 73% among experts but only 23% among the general public.