[BUG] [Cowork] Korean text garbled output with Sonnet model - vowel corruption and syllable dropping

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?

Summary

Cowork produces garbled Korean text when using the Sonnet model.
The output contains vowel corruption and syllable dropping that does not occur with Opus.

Environment

• Product: Claude Cowork
• Model: Sonnet (latest)
• OS: Windows
• Version: Latest (as of June 2026)

Reproduction Cases

The following Korean words were consistently output incorrectly:

| Expected | Actual | Error Type |
|----------|--------|------------|
| 솔루션사업팀 | 쎔루션사업팀 | Vowel corruption (ㅗ → ㅖ) |
| 빌딩의 신 | 빨딩의 신 | Vowel corruption (ㅣ → ㅏ) |
| 로그인 상태 | 로근 상태 | Syllable dropping (그인 → 근) |

Additional Notes

• Switching to Opus resolves the issue
• This is not an IME/input issue — the garbled text appears in Cowork's output
• Issue persists after updating to the latest version (post v2.1.116 fix)

Expected Behavior

Korean text should be output accurately regardless of model selection.

What Should Happen?

Korean text should be output accurately regardless of which model (Sonnet or Opus) is selected.

Error Messages/Logs

Steps to Reproduce

1. Open Claude Cowork
2. Select Sonnet as the model
3. Start a session in Korean or with Korean context
4. Observe the output — Korean characters appear garbled (vowel corruption, syllable dropping)

Claude Model

None

Is this a regression?

Yes, this worked in a previous version

Last Working Version

_No response_

Claude Code Version

1.10628.2

Platform

Anthropic API

Operating System

Windows

Terminal/Shell

Other

Additional Information

_No response_