Saltar para o conteúdo principal

Reasoning Configuration

Você pode configure o reasoning behavior using the reasoning parameter:

Reasoning Intensity

The effort parameter controls how much computational resource the model invests in reasoning, essentially dictating the level of reasoning effort.

Using Reasoning in Conversations

The reasoning feature can also be utilized in multi-turn dialogues:

Responses Containing Reasoning Information

When reasoning is enabled, the API returns results that include reasoning data:

Usage Recommendations

  1. Escolha the appropriate reasoning effort level: Use high for complex issues and low for simple tasks.
  2. Consider token usage: Reasoning will increase token consumption.
  3. Utilize streaming: For longer reasoning chains, streaming can enhance user experience.
  4. Provide context: Give the model sufficient context to enable effective reasoning.

Última atualização: 2026-06-01