claude-code - 💡(How to fix) Fix [BUG] Using the same prompt and model at different times results in higher token consumption during later executions. [2 comments, 3 participants]

Error Messages/Logs

TL;DR

The issue is likely related to a regression in version 2.1.107 of Claude Code, and the user should try to identify the last working version to help pinpoint the cause.

Guidance

• Since the issue is reported as a regression, comparing changes between the current version (2.1.107) and the last known working version could help identify the root cause.
• The user should attempt to recall or find the last working version of Claude Code to provide more specific information for debugging.
• Given the lack of specific error messages or logs, reproducing the issue with detailed logging enabled might provide crucial insights into what's going wrong.
• Checking the release notes or changelog for version 2.1.107 and subsequent versions might highlight changes that could be related to the regression.

Notes

The provided information lacks specific details about the error or the changes made in the recent version, making it challenging to provide a precise fix. More details about the last working version and any error messages would be necessary for a more targeted approach.

Recommendation

Apply workaround: The user should focus on identifying the last working version and comparing it with the current version to find a potential workaround or fix for the regression.