[BUG] Thinking messages not displayed in headless mode (-p flag) in 2.0.76 with OAuth

Preflight Checklist

• [x] I have searched existing issues and this hasn't been reported yet
• [x] This is a single bug report (please file separate reports for different bugs)
• [x] I am using the latest version of Claude Code

What's Wrong?

Thinking blocks are not being generated when using -p (print/headless) mode, even though thinking is enabled and works correctly in interactive mode with the same model (Opus 4.5), when signed in using OAuth (did not test API key), using v2.0.76.

What Should Happen?

Session file should contain "type":"thinking" blocks, similar to interactive mode sessions.

Actual behavior:

• Interactive mode: Thinking blocks present ✓
• -p mode: 0 thinking blocks ✗

Same model, same prompt, different behavior.

Error Messages/Logs

N/A

Steps to Reproduce

1. Confirm thinking is enabled in interactive mode via /config
2. Verify thinking works in interactive mode (establishes baseline):

``bash
claude
# Then enter: "Tell me a funny original joke about debugging code"
# Exit and check the session file:
ls -lt ~/.claude/projects/<project-slug>/*.jsonl | head -1
grep -c '"type":"thinking"' <newest-session-file>
# Should return 1 or more
``

1. Run the same prompt in -p mode:

``bash
claude -p "Tell me a funny original joke about debugging code"
``

1. Check the session file for thinking blocks:

``bash
ls -lt ~/.claude/projects/<project-slug>/*.jsonl | head -1
grep -c '"type":"thinking"' <newest-session-file>
# Returns 0
``

Claude Model

Opus

Is this a regression?

Yes, this worked in a previous version

Last Working Version

1.0.119

Claude Code Version

2.0.76 (Claude Code)

Platform

Anthropic API

Operating System

macOS

Terminal/Shell

iTerm2

Additional Information

• Thinking works correctly in interactive CLI mode with the same prompts
• Previously reported as a bug between 1.0.58 and 1.0.119, reported here: https://github.com/anthropics/claude-code/issues/7840
• The changelog for 1.0.119 mentioned "Fix thinking mode not working in headless sessions" - this may be a regression
• Attempted workarounds that did not work:
• --settings '{"thinkingMode": "enabled"}'
• --settings '{"thinking": {"enabled": true}}'
• --model opus explicitly
• --betas extended-thinking (returns "Custom betas are only available for API key users")