Simple, I have a large 3,000+ line script and ask it to make an assessment about the script. If “auto” is selected it fails and throws this message that its too long.
Yes because some models have different context sizes and auto selects the most capable regular model which is most available, instead of always a specific one.
It might be a good feature request instead of a bug report if you were to suggest this should be automatically handled. Note that the Cursor team has many cases and combinations to consider so every feature might not have a full spread of options etc baked in.
The term “auto” implies it would automatically select the correct the model for the job. And it tries to select a model that isn’t suited for the task, that seems like a bug to me.
In this case Auto means automatic, which Cursor defined in their docu and explanation in the forum, as in a model that is not overloaded at the moment. Not in any other interpretation of the word.
Simple, I have a large 3,000+ line script and ask it to make an assessment about the script. If “auto” is selected it fails and throws this message that its too long.
I have both a Mac book and a Ubuntu machine and this behavior seems to be on both.
If I manually select gemini and it works without an issue.