Unverified50% confidenceFactExact time
OpenAI o1、o3以及Anthropic Claude 3.7 Sonnet等推理模型会在返回答案前进行大量内部反思，中间推理Token对用户不可见但仍计入计费
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
7/21/2026
First Seen
Valid until: 10/19/2026
Sources
Related Claims
VerifiedO3属于OpenAI的o系列推理模型，引入了思维链（Chain-of-Thought）机制进行多步内部推理79% similarUnverified推理强度机制的慢思考范式源自 OpenAI o1 系列，Anthropic 随后在 Claude 3.7 及更新模型中跟进实现75% similarUnverified推理增强型模型（如o1、DeepSeek-R1）的思维链过程会消耗大量对用户不可见的推理token，o1-preview实际消耗token往往是可见输出的3-8倍75% similarUnverified推理模型（如OpenAI的o1/o3系列、Anthropic的Claude with extended thinking、DeepSeek-R1）在生成最终答案前会先输出思考过程（thinking tokens）75% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/579931API
curl https://kongchang.com/api/v1/knowledge/claims/579931MCP
get_claim(id=579931)