Token Limit Reached Extremely Fast Despite Low Context Usage

Resolved 💬 3 comments Opened Apr 25, 2026 by peco-tanaka Closed Apr 29, 2026

Bug Description
The tokens were consumed almost instantly—after just two prompts. Even though the instructions were given within a conversation using less than 20% of the context, it still hit the limit very quickly.

This issue has been occurring frequently even after reporting the bug fix, which is quite problematic.

It seems to happen when starting a conversation from a continued session.

Environment Info

  • Platform: darwin
  • Terminal: vscode
  • Version: 2.1.119
  • Feedback ID: 6e410768-6222-4200-a84d-01a94369ec19

Note: Content was truncated.

View original on GitHub ↗

This issue has 3 comments on GitHub. Read the full discussion on GitHub ↗