After my included usage reached 100%, I can no longer use Auto or Composer 2.5.
The banner says:
“You’re out of usage. Switch to Auto, Auto, or Composer 2.5…”
However, switching to Auto does not work. I also tried selecting Auto manually from the model dropdown, but it still shows that I am out of usage and I cannot continue using Cursor.
This appears to be the same issue discussed in the Cursor forum:
Steps to Reproduce
Use Cursor until the included usage reaches 100%.
Open a chat or Composer.
Switch to Auto (both from the banner and manually from the model dropdown).
Send a prompt.
Expected Behavior
Even after the included usage reaches 100%, Auto should still work (or at least it should work when selected manually if this is the intended fallback).
Instead, Cursor reports that I am out of usage and does not allow me to continue.
・This seems to match the known issue acknowledged by Cursor staff in forum topic #166442.
・I also confirmed reports that the same issue occurs on macOS, so it does not appear to be Windows-specific.
I encountered the usage limit after submitting only two prompts, and unfortunately the limit has still not reset after waiting for two days. This has prevented me from continuing my work and has made it difficult to evaluate the platform for a real development project.
I’m currently building a fairly large application, so reliability and uninterrupted workflow are very important. I’ve been using Antigravity for similar work and haven’t experienced this type of issue, which is why I was exploring this platform as a potential alternative and was even considering upgrading to the Pro plan.
Hey ■■■■■■■■■■■■■■■■,
You’ve hit a real bug on our side, not a problem with your account or your team’s settings. After you reach 100% of included usage, Auto and Composer 2.5 are meant to keep working, but the picker’s default variants (“Auto Balance” and “Composer 2.5 Fast”) aren’t being recognized as the allowed fallback, so they get blocked. It’s a known issue we’ve reported internally and are tracking.
Two reliable ways to unblock yourself right now:
Auto Cost. In the model picker, select Auto, then set its option to Cost (Auto has Cost / Balance / Intelligence variants, and the default is Balance). Auto Cost is the one your fallback accepts.
Plain Composer 2.5. Select Composer 2.5 and turn the Fast toggle off - the non-Fast version goes through fine.
One thing to avoid for now: the “Switch to auto” button in the error banner sets an invalid value and can throw a separate “model not found” error, so pick the model yourself from the picker instead. And if you’re a team admin (or can ask yours), raising the usage limit unblocks everyone right away.
@sonpn20_vsffin - same bug, and the important detail is that it has to be Auto Cost specifically, not Auto Balance. If you’ve selected Cost and it’s still blocked, could you grab a request ID (three-dots menu on the message → Copy Request ID) and a screenshot of what your Auto options show? I’ll trace your specific case.
@vivek_nigam - the steps above should get you working again immediately. If your limit genuinely isn’t resetting on your billing cycle (that would be separate from this bug), open a new thread with your account email and we’ll look into it directly.
Happy to point you to the exact spot for Auto Cost if it’s not obvious in your picker - a few people have found it a bit tucked away.
thanks for the request ID! I traced it, and your situation is a bit different from the original bug in this thread. You did pick Auto Cost correctly, so that part’s right - the problem is that your request is being blocked because your team has reached its usage limit, not by the model-picker issue the others here are hitting.
Since you’re on a team, the unblock is on the admin side: your team admin can raise the team’s usage limit (or turn on additional usage) in the dashboard and you’ll be able to continue right away. It’ll also free up when your team’s usage resets at the start of your next billing cycle.
If your admin bumps the limit and Auto Cost still won’t run, drop a fresh request ID here and I’ll take another look.
