[BUG] messages.37.content.0.thinking.valid: Extra inputs are not permitted

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?

The Claude Code session became completely unresponsive with repeated API 400 errors. Every subsequent user input triggered the same error, making recovery impossible.

Starting new conversation help to fix the issue but it happens randomly with few users.

Downgrading to v2.0.36 works. However, upgrading to v2.1.12 also work randomly

What Should Happen?

Expected Behavior:

Claude Code should handle long conversations gracefully without entering unrecoverable error states
When errors occur, the system should offer graceful degradation options

Error Messages/Logs

400 BAD_REQUEST - {"type":"error","error":{"type":"invalid_request_error","message":"messages.37.content.0.thinking.valid: Extra inputs are not permitted"},"request_id":"req_vrtx_011CXJnDJBrjyfHm6oYBvpwY"}

Steps to Reproduce

1. Steps to Reproduce seems to be difficult because Same issue is not happening for other team members with same prompts

Prompt 1 > Explore the codebase
Prompt 2 > Create a new folder and add your thoughts on how to improve this

Claude Model

Sonnet (default)

Is this a regression?

Yes, this worked in a previous version

Last Working Version

2.0.36

Claude Code Version

2.1.12

Platform

Google Vertex AI

Operating System

macOS

Terminal/Shell

Terminal.app (macOS)

Additional Information

Seem like in the above specs, “valid” should not come