DeepSeek models support thinking mode: before outputting the final answer, the model will first output a chain of thought content to improve the accuracy of the final answer. Thinking mode toggle and thinking intensity control Control parameters (OpenAI format) | Control parameters (Anthropic format) ---|--- Thinking mode toggle(1) | {"thinking": {"type": "enabled/disabled"}} Thinking intensity control(2)(3) | {"reasoning_effort": "high/max"} | {"output_config": {"effort": "high/max"}} (1) Default thinking toggle is enabled (2) low, medium will be mapped to high, xhigh will be mapped to max When using the OpenAI SDK to set the thinking parameter, you need to pass the thinking parameter into extra_body: response = client.chat.completions.create( model="deepseek-v4-pro", # ...

reasoning_effort="high", extra_body={"thinking": {"type": "enabled"}} ) Input and Output Parameters Thinking mode does not support temperature, top_p, presence_penalty, frequency_penalty parameters. Note that for compatibility with existing software, setting these parameters will not cause errors, but they will not take effect either. In thinking mode, the chain of thought content is returned through the…