Unverified50% confidenceFactExact time
Thinking models (such as Qwen 3.5, DeepSeek-R1, etc.) structurally use special markers (such as Thinking or ＜/thinking＞) to distinguish the internal reasoning process from the final answer, and the inference runtime needs to parse these markers correctly.
1
Sources
50%
Confidence
Long-term
Relevance
10/30/2026
First Seen
Sources
Related Entities
Related Claims
VerifiedReasoning models, represented by OpenAI o1, DeepSeek-R1, and QwQ, have chain-of-thought expansion as their core mechanism, generating a large number of intermediate reasoning steps before producing a final answer.79% similarVerifiedReasoning models (such as OpenAI's o1/o3 series, Anthropic's Claude with extended thinking, and DeepSeek-R1) output their thinking process (thinking tokens) before generating the final answer.78% similarVerifiedQwen3 27B has thinking mode enabled by default, supports disabling thinking, and controls reasoning intensity via a Reasoning Effort parameter.77% similarUnverifiedIn deep thinking mode, the model first generates intermediate reasoning steps internally, then derives the final conclusion based on those steps, performing significantly better than direct generation when handling multi-step logical reasoning tasks.77% similarUnverified一些推理增强型模型（如DeepSeek-R1、Kimi K3）会生成非常长的思维链，有时推理过程的Token数量是最终答案的数十倍76% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1115136API
curl https://kongchang.com/api/v1/knowledge/claims/1115136MCP
get_claim(id=1115136)