Unverified50% confidenceBenchmarkExact time
在科幻小说《极光》飞船复刻测试中，DeepSeek V4 Pro识别出搜索结果矛盾并主动更换关键词校验，表现反超Kimi K3
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
8/19/2026
First Seen
Valid until: 11/17/2026
Sources
Related Claims
UnverifiedIn testing, DeepSeek V4 Pro completed the basic functions normally, but had a dropdown style issue: only 3 of the 5 languages were visible and could not be scrolled through.68% similarUnverifiedIn a frontend page generation test, DeepSeek V4 achieved a high level of completeness in a single generation and correctly implemented the Easter egg feature.67% similarVerifiedDeepSeek V4 Pro adopts a MoE (Mixture of Experts) architecture, activating only a subset of parameters during inference to reduce computational costs.67% similarUnverifiedDeepSeek is suspected to be gray-testing a new V4 model, evidenced by a new chain-of-thought format appearing in its web interface quick mode65% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/774578API
curl https://kongchang.com/api/v1/knowledge/claims/774578MCP
get_claim(id=774578)