NIST CAISI: DeepSeek V4 Pro Trails US AI Models by 8 Months
The National Institute of Standards and Technology's CAISI evaluation has found DeepSeek V4 Pro to be the most capable Chinese AI model evaluated to date, but trailing leading US AI models by approximately eight months on capability benchmarks. This is the first official US government quantification of the US-China AI capability gap using a standardized evaluation framework. The CAISI (Comprehensive AI Systems Intelligence) benchmark covers reasoning, coding, and multimodal tasks.
Why It Matters
A government-issued capability gap estimate gives policymakers a concrete — and publicly defensible — metric for AI competitiveness debates, replacing informal industry comparisons with an official benchmark timeline.