Stanford University’s Human-Centered Artificial Intelligence Institute released the official Chinese version of its 2026 AI Index Report on Tuesday. According to ChainCatcher, the report says artificial intelligence is moving beyond a pure model-capability race and into broader changes in infrastructure, organizational processes, and institutional systems.
The report said leading models are now approaching human-level performance on the SWE-bench software engineering benchmark, while agents’ task success rate in real computer environments has risen from 12% to 66%. It added that AI is shifting from generating information to executing tasks.
The report also said the gap between top U.S. and Chinese models has narrowed to 2.7%, with the two sides’ leading models repeatedly swapping the lead. It further noted that physical infrastructure has become a core constraint on AI development, with global AI data center power capacity reaching 29.6 gigawatts, and chips, energy, and water emerging as key strategic bottlenecks.