general family · filed under inference & serving
Chain-of-thought reasoning
The model produces reasoning content before its answer; the evidence also describes full access to that reasoning in some cases.
Also called Chain-of-Thought, COT, deep chain-of-thought, reasoning tokens, Chain-of-thought reasoning mode, Full chain-of-thought.
- sources
- 8
- models
- 6
- labs adopt it
- 5
- strongest
- core
How sources treat it
One count per evidence span, weakest treatment to strongest.
Documented in
Further reading
Picked by hand, not extracted: where to read more, not evidence for anything on this page.
Evidence
8 spans quoted from the sources, strongest treatment first.
Full chain-of-thought: Gain complete access to the model’s reasoning process, facilitating easier debugging and increased trust in outputs.
Reasoning mode: Defaults to "high" (deep chain-of-thought), which suits complex tasks such as math, coding, and reasoning.
It only has one reasoning effort right now, 'max'—and it shows. The model consumed 13,241 reasoning tokens to output 3,417 tokens of response.
It uses extended thinking or chain-of-thought reasoning to work through complex problems before providing an answer.
Provides complete access to the model's reasoning process, facilitating easier debugging and greater trust in outputs.
generate both COT and TIR solution trajectories with DeepSeek-V4-Pro
reasoning_effort: "no_think" (default), "low", "high" (deep chain-of-thought)
reasoning_effort: 'no_think' (default, direct response), 'low', 'high' (deep chain-of-thought)
Filed alongside
Other methods under inference & serving :: reasoning control.