[BUG] Claude Code for VS Code intermittently omits thinking summaries on claude-sonnet-5 despite showThinkingSummaries being enabled

Status Closed — duplicate
Reported on v2.1.228
Maintainer reply None cached
Activity 1 comment · opened Aug 12, 2026 · closed Aug 15, 2026

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?

Claude Code for VS Code shows inconsistent thinking-summary behavior on claude-sonnet-5. Some turns show a visible written thinking summary, while other turns in the same setup return a thinking block with an empty thinking field and only a signature. This affects day-to-day work because the visible workflow / reasoning summary is sometimes missing even when local display settings are configured correctly.

What Should Happen?

If Claude Code is configured to show thinking summaries, the extension should behave consistently: every eligible turn should either show summarized thinking text, or the product should expose a documented setting that clearly controls whether display: "summarized" is requested.

Error Messages/Logs

Relevant local observations from Wednesday, August 12, 2026:
- Claude settings enabled: showThinkingSummaries=true, alwaysThinkingEnabled=true, verbose=true, viewMode=verbose, defaultView=transcript
- VS Code setting enabled: claudeCode.focusView=false
- In local transcript files, some assistant messages had type=thinking with non-empty text
- Other assistant messages had type=thinking with an empty thinking field and only a signature
- Attempted model override to claude-sonnet-4-20250514 failed with: "There's an issue with the selected model (claude-sonnet-4-20250514). It may not exist or you may not have access to it."

Steps to Reproduce

  1. In Claude Code for VS Code, enable showThinkingSummaries and verbose output, and ensure Focus View is off.
  2. Use claude-sonnet-5 in an interactive session.
  3. Run normal prompts that trigger tool use or visible thinking.
  4. Observe that some turns show a written thinking summary while other turns show only "Thought for ..." with no written summary text.
  5. Check the local Claude transcript file and compare assistant messages: some contain non-empty thinking text, while others contain an empty thinking field plus signature only.

Claude Model

None

Is this a regression?

Yes, this worked in a previous version

Last Working Version

Unknown

Claude Code Version

2.1.228

Platform

Anthropic API

Operating System

macOS

Terminal/Shell

Terminal.app (macOS)

Additional Information

Environment:

  • Product: Claude Code for VS Code
  • Extension version: 2.1.228
  • VS Code version: 1.133.0
  • Model path observed in transcript: claude-sonnet-5
  • OS: macOS

Workarounds tried:

  • Confirmed local settings were correct
  • Reloaded the VS Code window
  • Confirmed the extension was up to date
  • Tried switching to claude-sonnet-4-20250514 as a user-level model override, but that model was not available on this account

Anthropic support indicated filing on GitHub could provide additional visibility.

View original on GitHub ↗

This issue has 1 comment on GitHub. Read the full discussion on GitHub ↗