Within the last hour, Gemini 2.4 Pro Max stopped working. I have an Google API key that I’m using and the workflow has been working great since yesterday. I took a break from coding, attended a meeting, came back and it was broken.
Troubleshooting steps:
I’m more so filing this for support visibility - I don’t think I’m a unique case.
Same problem cursor just dont want we use api and get more smart models for lower price!
Same issue - very frustrating the number of problems with API keys. Continued failures with Gemini and still no support for AWS/Bedrock makes Cursor borderline unusable.
Hey, it seems like you’re trying to use the Gemini Max model. This is our custom model specifically developed for use in Cursor. Naturally, it’s not available in the Google API, which is why you’re encountering an error about a non-existent model. You should use gemini-2.5-pro-exp-03-25 instead.
Naturally, I would like to know how I can use the custom model specifically developed for use in Cursor, which was what I was using up until about an hour ago until it started returning 404’s
You just need to disable your API key if you want the model from your Cursor subscription.
how was it working before the change though. oh my god was it charging me this whole time
I was just following what danperks suggested in a different thread:
As I mentioned, I would recommend using your own API keys while Google offers this, and with a Pro subscription, everything should work as expected here. Max mode is also free when using an API key, and uses all the context available!
Gemini 2,5 performance: Great yesterday, terrible today - #30 by danperks
It was working fine all day and then suddenly stopped, as if an update had been pushed (either front end or backend ¯_(ツ)_/¯ ) - so I’m not sure how you can say it’s not available in the Google API. Seems to conflict with other information out there.
the same thing with the NE max model there are much fewer tokens maximum! lmao how do they want to save money
Has anyone found a solution ? I still can not use 2.5 Pro Max with my Gemini API Key…
This doesn’t make any sense - what do you mean “custom model”?
You’re not training/fine-tuning your own version of Gemini, and my understanding was that this was related to the context length sent with requests.
As others have pointed out, this also doesn’t explain why this worked perfectly fine up until this past week.
There have been numerous communications from the Cursor team saying that Max was available at cost with bring-your-own API key.
If this is meant to force people into paying extra, then you should please just say that. It would be disappointing and reflect poorly on Cursor, but would at least clear up some of the confusion. Right now you’re forcing users to waste time troubleshooting, providing no/incorrect support, and extracting money from us for a feature that worked without issue since it was released (if you can call loosened context restrictions a feature).
The other evidence that suggests some kind of dishonesty here is that Anthropic API keys seem to work perfectly fine with the Claude version of Max. If Max is some type of “custom” model as you say, then I would assume this would also not work?
I really don’t want to go through the time and hassle of switching off of cursor and trying alternatives, but this is both frustrating and leaves a very bad taste in my mouth.
I saw that your team said on the forums that gemini pro max is compatible and that it worked fine for the first couple of days, so why is it not working since yesterday?
This whole thing is confusing. I’ve been charged for MAX calls all month without knowing, because I didn’t put in my API key. Why didn’t I put my key in? I was told all google API calls would be made using my pro subscription…
Now it seems like it’s actually impossible to charge MAX calls to my pro subscription OR API key?!
Max is a custom model? No, we’re paying for inference, not some prompt you guys have tacked onto the model. That’s what we pay our pro subscription for.
I’d rather unsubscribe and use alternative tooling thanks.
Hey guys, just want to set the record straight, and clarify a few things!
Max mode is not a custom trained model, but is a mode you can enable for certain models which allows you to use the entire context window and more tool calls in one go. For older models, too much context would cause them to perform poorly, but newer Max-supported models work better with this, hence the reason for the feature to be added.
You should definitely be able to use Max mode with an API key, where you won’t be charged anything to your Cursor account. You do still need a Pro account, as Cursor still uses smaller models under the hood for certain purposes, but any Max specific costs would go to your own API key if one is entered.
Thanks for that clarification, @danperks. It sounds like it’s just an unfortunate clashing of names, one for the mode and one for the model name, that caused the confusion on my end. I appreciate the response. Cheers!
Actually, I’m still confused.
From what I can see here:
It looks like Gemini 2.5 Pro, MAX mode is something that should be supported. However, when I select gemini-2.5-pro max in the … I get this error:
Request failed with status code 404: [{
"error": {
"code": 404,
"message": "models/gemini-2.5-pro-max is not found for API version v1main, or is not supported for generateContent. Call ListModels to see the list of available models and their supported methods.",
"status": "NOT_FOUND"
}
}
]
I haven’t put in any custom models or anything like that. gemini-2.5-pro-max is one of the available models from Cursor… but when I select it I get an error.
What am I missing?
Also, when I disable my Google API Key and then try the Gemini 2.5 Pro Max mode, it works. So… something about using the Google API Key is conflicting with the ability to use Max mode with the Gemini 2.5 Pro model, contrary to what @danperks is suggesting 
