Leverage Adaptive Thinking for Complex Logical Reasoning
Let the model automatically scale its reasoning for complex logic.
Trust the new adaptive thinking feature to handle complex reasoning automatically without needing to manually toggle 'thinking' modes or 'speed' settings.
The Scenario
You are designing a complex logic flow for a business process that has multiple conflicting rules. You need the AI to think through the implications of each rule to ensure no contradictions exist.
Before & after
Users manually toggled 'extended thinking' modes or wrote complex 'Chain of Thought' instructions. Setting up and refining these prompts took 10–15 minutes.
Engage the model with a complex prompt. In 1–2 minutes, Claude Sonnet 5 uses adaptive reasoning to internally verify its logic before giving a highly accurate response.
The Prompt
[INSERT_COMPLEX_PROBLEM]. Solve this step-by-step, ensuring you verify the edge cases for [SPECIFIC_CONSTRAINT]. If you identify an error in your initial logic, correct it before providing the final answer.
Sonnet 5 enables adaptive thinking by default. This means the model automatically determines how much 'thought' or reasoning energy to apply to a problem, making it highly effective for agentic tasks and coding.
Source
Claude Platform - Claude Platform Docs"adaptive thinking is now on by default; manual extended thinking... is removed."