Expired50% confidenceBenchmarkExact time
In actual testing, the cache hit rate of DeepSeek Harness executing multi-step programming tasks reached over 99%, with four tasks costing less than 5 yuan in total.
1
Sources
50%
Confidence
Short-term (~14 days)
Relevance
9/7/2026
First Seen
This claim has expired. It is kept for the record and no longer describes the current state.
Valid until: 9/21/2026(expired)
Sources
Related Entities
Related Claims
UnverifiedDeepSeek's inference cost per unit of performance is approximately 1/5 to 1/10 that of dense models at the same tier.72% similarUnverifiedDeepSeek replicated top-tier model performance at extremely low compute costs.70% similarUnverifiedDeepSeek's training cost is only a fraction of that of comparable models.69% similarUnverifiedDeepSeek的训练成本仅为同等性能国际模型的几分之一，API调用价格远低于OpenAI和Anthropic的产品68% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/870128API
curl https://kongchang.com/api/v1/knowledge/claims/870128MCP
get_claim(id=870128)