Unverified50% confidenceFactExact time
On Opus 5.5, asking Claude to write out its full reasoning process before answering may be flagged and rejected, with the error reason shown as 'reasoning extraction,' intended to prevent other companies from distilling Opus 5.5.
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
10/7/2026
First Seen
Valid until: 1/5/2027
Sources
Related Entities
Related Claims
UnverifiedIn long conversations, Opus 5.5 sometimes goes back to reconsider previous answers, and this can be addressed by using project instructions to tell it to treat already-answered content as finalized in order to maintain focus.74% similarUnverifiedClaude Opus 5.5's thinking process is always on and cannot be fully disabled; to save tokens, the effort level should be lowered rather than thinking turned off.71% similarUnverifiedOpus 5对高风险请求可能被分类器拦截或触发回退机制71% similarUnverifiedThe most noticeable writing characteristic in Opus 5.5 model output is an abnormal preference for the word 'dependable'.68% similarUnverifiedReasoning models exhibit overthinking at excessively high effort levels, causing them to overturn originally correct answers, and a similar pattern has been observed in Anthropic's Opus 5.67% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/984280API
curl https://kongchang.com/api/v1/knowledge/claims/984280MCP
get_claim(id=984280)