Switch Between Thinking and Non-Thinking Modes via Model Aliases
Select the right model alias to balance speed and logic.
Choose 'deepseek-reasoner' for tasks requiring intense thought and 'deepseek-chat' for fast, conversational responses.
The Scenario
You need to solve a difficult math problem or a complex coding challenge that requires a 'Chain of Thought' approach rather than a quick conversation.
Before & after
Users manually write long, complex prompts to force a standard model to 'think step-by-step,' which can take 5-10 minutes to draft and refine.
Quickly switch between 'deepseek-chat' for speed and 'deepseek-reasoner' for logic. The AI provides a refined answer in under 1 minute.
The Prompt
[ACTIVATE_REASONER_MODE] Solve the following complex math and logic puzzle step-by-step: [PASTE_PROBLEM_HERE]
The DeepSeek API uses two distinct model aliases: 'deepseek-chat' for non-thinking (regular chat) and 'deepseek-reasoner' for thinking mode (extensive logic).
Source
Change Log | DeepSeek API Docs"deepseek-chat corresponds to DeepSeek-V3.2's non-thinking mode; deepseek-reasoner corresponds to DeepSeek-V3.2's thinking mode."