Extended Thinking

Extended Thinking gives Claude enhanced reasoning capabilities for complex tasks, while providing varying levels of transparency into its step-by-step thought process before it delivers its final answer.

Supported models

Extended thinking is supported in the following models:

• Claude Sonnet 4.5 (claude-sonnet-4-5-20250929)
• Claude Sonnet 4 (claude-sonnet-4-20250514)
• Claude Sonnet 3.7 (claude-3-7-sonnet-20250219)
• Claude Haiku 4.5 (claude-haiku-4-5-20251001)
• Claude Opus 4.1 (claude-opus-4-1-20250805)
• Claude Opus 4 (claude-opus-4-20250514)

How extended thinking works

When extended thinking is turned on, Claude creates thinking content blocks where it outputs its internal reasoning. Claude incorporates insights from this reasoning before crafting a final response.

The API response will include thinking content blocks, followed by text content blocks.

Response Format Example

{
  "content": [
    {
      "type": "thinking",
      "thinking": "Let me analyze this step by step...",
      "signature": "WaUjzkypQ2mUEVM36O2TxuC06KN8xyfbJwyem2dw3URve/op91XWHOEBLLqIOMfFG/UvLEczmEsUjavL...."
    },
    {
      "type": "text",
      "text": "Based on my analysis..."
    }
  ]
}

How to use extended thinking

Basic Usage Example

import anthropic

client = anthropic.Anthropic()

response = client.messages.create(
    model="claude-sonnet-4-5",
    max_tokens=16000,
    thinking={
        "type": "enabled",
        "budget_tokens": 10000
    },
    messages=[
        {
            "role": "user",
            "content": "Are there an infinite number of prime numbers such that n mod 4 == 3?"
        }
    ]
)

Key Parameters

Use Cases

Extended thinking is particularly well-suited for:

1. Complex Problem Solving

Mathematical proofs, logical reasoning, algorithm design, and other tasks requiring deep thinking.

# Example: Mathematical problem
message = client.messages.create(
    model="claude-sonnet-4-5",
    max_tokens=16000,
    thinking={"type": "enabled", "budget_tokens": 10000},
    messages=[{
        "role": "user",
        "content": "Prove: There are infinitely many primes p such that p ≡ 3 (mod 4)"
    }]
)

2. Architecture Design

System design tasks that require weighing multiple options and considering various constraints.

3. Debugging Complex Issues

Bug debugging that requires analyzing multi-level error causes and tracking complex dependencies.

4. Strategic Planning

Decision tasks that require multi-step reasoning and evaluating different paths.

Best Practices

1. Set Appropriate Token Budget

# Simple problems: smaller budget
thinking={"type": "enabled", "budget_tokens": 2000}

# Complex problems: larger budget
thinking={"type": "enabled", "budget_tokens": 10000}

2. Process Thinking Content

for block in response.content:
    if block.type == "thinking":
        # Optional: log thinking process for debugging
        print(f"Thinking: {block.thinking}")
    elif block.type == "text":
        # Final answer
        print(f"Answer: {block.text}")

3. Verify Thinking Authenticity

Extended thinking responses include digital signatures for verifying content authenticity:

thinking_block = response.content[0]
signature = thinking_block.signature
# Use Anthropic's verification method
is_authentic = anthropic.verify_thinking(thinking_block.thinking, signature)

Pricing Information

When using extended thinking:

• Thinking tokens: Billed at output token price
• Budget control: Use budget_tokens to control thinking costs
• Total cost: = Input token cost + Thinking token cost + Output token cost

Combining with Other Features

With Prompt Caching

response = client.messages.create(
    model="claude-sonnet-4-5",
    max_tokens=16000,
    thinking={"type": "enabled", "budget_tokens": 10000},
    system=[
        {
            "type": "text",
            "text": "You are a mathematics expert...",
            "cache_control": {"type": "ephemeral"}
        }
    ],
    messages=[...]
)

With Tool Use

Extended thinking can help Claude better plan and execute tool calls:

response = client.messages.create(
    model="claude-sonnet-4-5",
    max_tokens=16000,
    thinking={"type": "enabled", "budget_tokens": 5000},
    tools=[...],
    messages=[...]
)

FAQ

Q: When should I use extended thinking?

A: When tasks require:

• Multi-step reasoning
• Weighing multiple options
• Complex logical analysis
• Deep problem solving

For simple queries or creative writing, extended thinking is typically not needed.

Q: How to balance thinking quality and cost?

A:

• Start with smaller budget_tokens for testing
• Adjust budget based on task complexity
• Monitor actual thinking content usage
• Perform cost analysis for batch tasks

Q: Is thinking content reliable?

A:

• Thinking content includes digital signatures for authenticity verification
• The thinking process reflects the model’s actual reasoning
• However, thinking content may still contain errors; evaluate alongside final answer

Related Resources

• Claude API Documentation
• Prompt Engineering Guide
• Tool Use Documentation
• Model Context Protocol

Next Steps

• Explore Computer Use to let Claude control desktop environments
• Learn about Agent Skills to build reusable skill systems
• Check out Cookbook for more code examples