Free Edition doesn't publish an exact GPU quota number like "X GPU hours per day." What you're hitting is Databricks' fair use throttling: once your workspace crosses its usage threshold, compute gets shut off for the rest of the day, and in harsher cases for the rest of the month, then comes back automatically once the limit resets. Your notebooks and data stay intact the whole time, it's only compute that pauses.

Given you only ran a GPU cluster for 5 minutes, a couple things could explain it:

If this is the first time you've hit the block, it's most likely just the daily reset. Wait a day and try attaching again.

Serverless GPU on Free Edition is also capacity constrained on top of your own usage. Databricks' own docs say it's "subject to available capacity," so you can see a quota style error even without burning much usage yourself, if the shared GPU pool is busy right then.