Stanford AI Index 2026: China Has 'Nearly Erased' US AI Lead — Gap Shrinks from 1,300 to 39 Arena Points

Stanford University's Human-Centered AI Institute released its landmark 2026 AI Index report on April 16, delivering a striking finding: China has 'nearly erased' the United States' artificial intelligence performance lead. In May 2023, GPT-4 led China's best models by more than 1,300 points on the LMSYS Chatbot Arena benchmark. By March 2026, that gap had shrunk to just 39 points — Anthropic's Claude Opus 4.6 leads China's Dola-Seed 2.0 by only 2.7%. Additional findings: China surpassed the US in AI research publication citation share (20.6% vs. 12.6%); China leads in industrial robotics deployment (~295,000 installations vs. US 34,200); the flow of AI scholars moving to the US dropped 89% since 2017, with an 80% decline in just the past year. DeepSeek's R1 and V3 models were specifically cited as evidence of China's accelerating competitive position. The Stanford findings directly challenged the US national security community's assumption that export controls are effectively widening the US-China AI performance gap — the data indicates the gap has narrowed dramatically despite escalating chip sanctions.