Expired85% confidenceFactTime unknown
DeepSeek V4 Pro's floating-point operations have been reduced to 27% of the V3.2 version (a 3.7x reduction).
1
Sources
85%
Confidence
Medium-term (~90 days)
Relevance
6/2/2026
First Seen
This claim has expired. It is kept for the record and no longer describes the current state.
Valid until: 8/31/2026(expired)
Sources
Related Entities
Related Claims
UnverifiedDeepSeek V4 Pro achieved a score of 3206 on CodeForces and a pass rate of 95.2% on the MMT-266 benchmark.75% similarUnverifiedDeepSeek V4 Pro leads Qwen3.6 27B Q4 by approximately 6%, and if it had not unexpectedly underperformed on the reasoning-mathematics items, the lead could have exceeded 10%.73% similarUnverifiedLaguna S 2.1 claims to be "14× smaller and 4.5× stronger" in coding capability relative to large models such as DeepSeek V4 Pro Max69% similarUnverifiedBased on comparative testing, DeepSeek V4 Pro's programming performance even surpasses Claude Sonnet, and its price is much cheaper.68% similarUnverifiedQwen3.5's parameter activation rate is 4.3%, while DeepSeek V3's 671B model has an activation rate of 5.5%.66% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/36597API
curl https://kongchang.com/api/v1/knowledge/claims/36597MCP
get_claim(id=36597)