Context limit gauge percent bug

Where does the bug appear (feature/product)?

Cursor IDE

Describe the Bug

The stat for context used during a conversation is inaccurate. It fills up exceptionally quickly the start of a conversation, but then once it hits 100%, it pulls back down to a more reasonable number like 17%, and then it climbs at a much more normal speed.

Steps to Reproduce

In the new agents window, when you hover over the chat, that’s where the percent of context taken up is, and it’s in the form of a gauge. Right now, for Opus 4.6, it gives you the illusion that you’re rapidly filling up the context, but once it hits 100%, it goes back down to 17% or so. I think there may be some kind of confusion about which model it is or some archaic system regarding how Opus used to have different costs at different context window ranges. It is distracting as I check my work to see how expensive my next message is.

Operating System

MacOS

Version Information

Version: 3.1.0-pre.12.patch.0
VSCode Version: 1.105.1
Commit: 157ffe1c6dac02720a7c5d875229565305625f10
Date: 2026-04-06T16:39:06.552Z
Layout: editor
Build Type: Stable
Release Track: Default
Electron: 39.8.1
Chromium: 142.0.7444.265
Node.js: 22.22.1
V8: 14.2.231.22-electron.0
OS: Darwin arm64 25.3.0

For AI issues: which model did you use?

Opus 4.6

Does this stop you from using Cursor

No - Cursor works, but with this issue

Hey, thanks for the detailed report. This looks like a bug in how the context gauge calculates the percentage early in a conversation. It’s probably getting the wrong max token value at first, so it jumps to 100%, then fixes itself once the correct value comes through.

I flagged this with the team. No timeline yet, but your report helps us prioritize it. Let me know if you notice anything else.