Unverified50% confidenceBenchmarkExact time
The Box team applied Opus 5.5 to enterprise knowledge work, achieving 63% token savings, 42% reduction in redundancy, and 30% speed improvement compared to Opus 5, with accuracy gains of 39% on financial services tasks and 65% on cloud cost analysis.
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
10/5/2026
First Seen
Valid until: 1/3/2027
Sources
Related Entities
Related Claims
Unverified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67% similarUnverifiedFor a task that sequentially calls 5 APIs, even if the success rate of each step is 95%, the overall success rate drops to approximately 77%.63% similarUnverifiedIn a test checking whether AI-modified code could be accepted by a project, Opus 5.5 using the default setting scored slightly higher than Astra's best publicly available score, and the cost per task was approximately one-fifth of Astra's.61% similarUnverifiedCompared to Claude Fable 5.1, Opus 5.5's per-task cost is reduced by 67.6%.61% similarUnverifiedThe developer used the medium reasoning tier of Opus 5.5 to complete the React frontend development, consuming only 35% of that week's usage quota.59% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/975202API
curl https://kongchang.com/api/v1/knowledge/claims/975202MCP
get_claim(id=975202)