Anthropic Extended Thinking
OpenAI Reasoning Models (o-series)
OpenAI’s o-series models reason natively — no configuration needed:o1, o3, o3-mini, o4-mini
Reasoning models don’t support
temperature, top_p, frequency_penalty, presence_penalty, or stop. These parameters are automatically stripped.Gemini Thinking
Gemini 2.5+ models support thinking with a token budget:xAI Reasoning Models
xAI Grok models with-reasoning in the name use chain-of-thought:
Thinking in Streaming
When usingprocessStream(), thinking content arrives as separate chunks:
Testing in Prompt Studio
Test thinking directly in the Portal without deploying an agent:- Go to Prompts and open any prompt
- Click the config icon (sliders) in the top bar
- Click Thinking to enable it (set budget tokens if needed)
- Send a message — the thinking content appears as a collapsible “Pensou sobre a resposta” block
With LLM Standalone
Provider Support
Next Steps
Structured Output
Get guaranteed JSON responses
Streaming
Real-time streaming responses