CAISI Evaluation Report Ranks DeepSeek V4 as China's Most Powerful Model, Trailing US Frontier by 8 Months

A new evaluation report from CAISI (Center for AI Safety and Innovation) ranks DeepSeek V4 as the most powerful AI model developed in China. The report, shared on Reddit and linked to a NIST.gov page, provides a detailed comparison of AI capabilities between Chinese and US models. DeepSeek V4 achieved top scores among Chinese models, but the report explicitly states it still trails US frontier models by roughly eight months. This timeline suggests that despite China's rapid progress, the US maintains a significant lead in AI development.

The report highlights specific metrics where DeepSeek V4 falls short, including reasoning, multi-turn conversation, and efficiency. The eight-month gap is measured against models like GPT-4o and Claude 3.5, which continue to advance. The findings underscore the difficulty of closing the AI gap, even for China's most advanced lab. For professionals, this means relying on US models for cutting-edge applications while Chinese models are strong but not yet competitive at the frontier.