Unverified80% confidenceFactTime unknown
DeepSeek V4 Pro在CodeForces取得3206分，MMT-266评测达到95.2%通过率
1
Sources
80%
Confidence
Medium-term (~90 days)
Relevance
5/31/2026
First Seen
Valid until: 8/29/2026
Sources
DeepSeek V4深度解析：混合注意力+流形约束+MOM优化器三大创新全解读
bilibili赋范大模型课堂
Related Entities
Related Claims
UnverifiedDeepSeek V4 Pro的浮点运算量降至V3.2版本的27%（降低3.7倍）75% similarUnverified在LiveBench等代码相关任务上，QwQ-32B超过DeepSeek R1约两个百分点69% similarUnverifiedDeepSeek V4 Pro在Humanity's Last Exam基准测试中得分37.768% similarUnverifiedOmniMerge V4在MBPP代码测试中达到73.4%，相比原版Q6K量化的56.2%提升了15.8个百分点67% similarUnverifiedDeepSeek V4 Pro相比Qwen3.6 27B Q4领先约6%，如果不是在推理数学项上意外翻车，领先幅度可能超过10%67% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/13865API
curl https://kongchang.com/api/v1/knowledge/claims/13865MCP
get_claim(id=13865)