Unverified50% confidenceBenchmarkExact time
在业务自动化评测中，相同单任务成本下Opus 5的通过率约为次优模型的1.5倍
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
8/23/2026
First Seen
Valid until: 11/21/2026
Sources
Related Entities
Related Claims
UnverifiedEach model is run five times at each inference effort level and the results are averaged.71% similarUnverifiedAnthropic released the Fable 5 model, which consumes twice as many tokens as Opus 4.8 during operation.69% similarUnverified自动微分在保持机器精度的同时，计算开销仅为前向传播的常数倍（通常2-5倍）69% similarUnverifiedBy request volume, the usage frequency of Cursor's agent is approximately 5 times that of assistive features such as Tab.68% similarUnverified多路径采样意味着数倍推理开销，若需5次独立采样成本会增加至少5倍67% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/789615API
curl https://kongchang.com/api/v1/knowledge/claims/789615MCP
get_claim(id=789615)