AI / Claude Models Basics Interview Questions
What is extended thinking and how does it differ from adaptive thinking in Claude?
Both features enable Claude to reason more carefully before answering, but they work differently and are available on different models.
| Feature | Extended Thinking | Adaptive Thinking |
|---|---|---|
| What it does | Claude produces explicit | Claude internally allocates more reasoning compute when a task requires it — no visible thinking blocks |
| Availability | Claude Haiku 4.5 only | Claude Opus 4.8, Sonnet 5, and Claude Fable 5 (always-on for Fable 5) |
| User control | Opt-in — you enable it with a parameter | Automatic on supported models; Fable 5 always uses it |
| Use case | When you want to see and verify the model's reasoning chain | General accuracy improvement, especially for complex tasks |
| Output impact | Adds thinking tokens to the response (billed separately) | No additional visible output |
# Enabling extended thinking on Claude Haiku 4.5 message = client.messages.create( model="claude-haiku-4-5-20251001", max_tokens=16000, thinking={ "type": "enabled", "budget_tokens": 10000 # max tokens for thinking }, messages=[{"role": "user", "content": "Solve: if 3x + 7 = 22, what is x?"}] ) # Response includes a "thinking" content block followed by the answer
Interleaved thinking (thinking between tool calls) is automatic on models with adaptive thinking and requires the interleaved-thinking-2025-05-14 beta header on earlier models like Opus 4.5 and Sonnet 4.5.
On which Claude model is Extended Thinking (visible reasoning blocks) available?
How does Adaptive Thinking differ from Extended Thinking in Claude?
More Related questions...
