The Stanford Institute for Human-Centered Artificial Intelligence (HAI) has released the ninth edition of its AI Index Report for 2026. This document stands as one of the most comprehensive overviews of artificial intelligence development, drawing on data from a wide range of sources.
A core finding of the report is that AI capabilities are not stagnating but are actually continuing to accelerate. The industry was responsible for releasing over 90% of notable frontier models in 2025. Some of these models are already meeting or exceeding human-level performance in PhD-level scientific tasks, multimodal reasoning, and competitive mathematics. On the SWE-bench Verified benchmark, performance scores jumped from 60% to nearly 100% in just a single year.
The report highlights a significant convergence between the United States and China. American and Chinese models have traded the lead several times since the start of 2025. In February 2025, DeepSeek-R1 briefly matched the top American model, and as of March 2026, Anthropic's leading model holds a narrow advantage of only 2.7%.
The U.S. remains the leader in the number of top-tier models and high-impact patents, while China leads in publication volume, citations, total patents, and industrial robot installations. South Korea stands out for its innovation density, leading the world in AI patents per capita.
What lies behind these figures, and how rapidly is the global landscape evolving?
The report also observes that responsible AI development is lagging behind capability growth, as safety benchmarks fall short and documented incidents rose to 362, up from 233 in 2024. Meanwhile, the U.S. maintains its dominance in private AI investment, totaling $285.9 billion in 2025—a figure more than 23 times higher than China's investment.
