Unverified50% confidenceBenchmarkExact time
On 36 small verification tasks, Opus 5.5 used about 40% fewer tokens than Opus at the halfway point.
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
10/24/2026
First Seen
Valid until: 1/22/2027
Sources
Anthropic Research: LLMs Make Formal Software Verification at Scale Possible
bilibilimo2cc10/24/2026
Related Entities
Related Claims
UnverifiedAnthropic states that Opus 5.5 is more than 30% faster than the previous generation and can save approximately 40% of compute under default settings.71% 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.69% similarUnverifiedAccording to data from Artificial Analysis, the number of reasoning tokens per task doubled from approximately 42,000 in Opus 5 to 84,000 in Opus 5.5, while output tokens only increased from 30,000 to 35,000.68% similarUnverifiedClaude Opus 5.5 achieved a perfect 20-out-of-20 score on only 241 of the 507 tasks, for a 20-of-20 reliability of 47.53%.68% similarUnverifiedSome 'high partial scores' in evaluations mask a true delivery rate of only 0% to 4%.68% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1092432API
curl https://kongchang.com/api/v1/knowledge/claims/1092432MCP
get_claim(id=1092432)