Claude Sonnet 5.5 thinks by default when thinking is omitted

Q: How do I fix this when moving to claude-sonnet-5-5?

Read content blocks by type (not content[0].text), pass thinking blocks back unchanged in tool loops, and revisit max_tokens, which covers thinking plus text. To keep running without up-front thinking, send thinking: {"type": "between_tools"}.

Behavior change Applies to claude-sonnet-5-5 on the Anthropic Messages API.

Anthropic's docs describe this behavior change but don't quote the exact message, so we don't either.

How to fix it

Read content blocks by type (not content[0].text), pass thinking blocks back unchanged in tool loops, and revisit max_tokens, which covers thinking plus text. To keep running without up-front thinking, send thinking: {"type": "between_tools"}.

Before

{
  "max_tokens": 1024,
  "messages": [
    {
      "content": "Summarize this ticket.",
      "role": "user"
    }
  ],
  "model": "claude-sonnet-4-5-20250929"
}

After (Claude Sonnet 5.5)

{
  "max_tokens": 1024,
  "messages": [
    {
      "content": "Summarize this ticket.",
      "role": "user"
    }
  ],
  "model": "claude-sonnet-5-5",
  "thinking": {
    "type": "adaptive"
  }
}

Check your whole request in the linter