First of all, I don’t even have an option for now as shown in the image below. Does anyone know how to fix this? I had it an update or 2 ago and it disappeared. This hadn’t mattered for me before but now that the new pricing seems to prefer auto mode specifically for unlimited usage, seems I’ll just get stuck with a bricked cursor once I hit my limit for the month.
Funny timing for my auto button to disappear right when auto gets unlimited usage compared to selecting a model so I’m not sure if this is just for me or for everyone but I’d love to hear if anyone found a fix.
Now the other questions I have, assuming I am able to find a way to get my auto mode back are:
Does anyone know if auto mode has any sort of hourly or monthly rate limit or does it seem to be more or less unlimited usage (within reasonable means) in anyones findings?
Does auto mode run using the selected user models or might it randomly use say claude-3.5-haiku even if I never want a response from that one?
Does auto mode attempt to intelligently use a model based on your request or is it just whichever is most available at the time?
Can you use max in auto mode for requests that should include longer context and if so does that effect whatever rate limiting is in place?
Does max mode on regular requests use up the $20 frontier model usage faster than simply the extra token usage or would you just use 200k worth of tokens instead of 120k worth?
If I haven’t used up my $20 of frontier usage yet, and use auto mode, does that first burn through my frontier model usage or can I save those for more important later tasks?
Are there still “premium” and non premium models where I can use say the gemini flash model to save my $20 of credits for expensive models when I need them or does every model now burn into that included limit if not using auto mode?
Does auto mode tell me what model was used when I use it?
If anyone has any thoughts or answers on any of these or related questions I’d love to hear them.
Unfortunately it seems that every time I return to finish a new project I have to relearn all of the nitty-gritty of the pricing scheme. To an extent I understand not having every one of these questions listed somewhere in the documentation since there are so many weird cases (and part of my issue this time may just be not even having the auto model option to test things), but hopefully this pricing model at least sticks around for a while (assuming it works well since I haven’t tested it much yet).
I used Auto to complete the first revision of a new project, and before starting with the prompt, I told it to inform me which model it would use (one that can successfully distinguish between GPT and Claude).
Based on the output and my judgment of the task’s effectiveness, it did use GPT to start a complex task, and the result was not good. In my opinion, it’s clearly a complex task that should have used Claude or Gemini Pro, but it didn’t. Now I’m very worried about Auto’s effectiveness.
I’ve been reading the forum for several days, all the messages. I sent a request to the support service. An explanation was created in the topic on how to return to the Opt Out options.
As a result, there is no answer anywhere - not in the support email, not on the forum, not in any of the topics out of nowhere. This is very strange, but everything is covered in fog. It’s a pity, but it seems that the decision has already been made for me (((
So I don’t know what they will answer your questions, it will be interesting. If they answer =)
I just checked and don’t have any api keys enabled. I found a few other people that seem to have the same issue in a bug report actually unfortunately there doesn’t seem to have been a solution found in a few days: “There is no option for the auto model”
Nope, Auto is totally unlimited and has a 0 cost on your rate limit
It will randomly choose between a selected set, including Claude 4, Gemini 2.5 Pro, GPT 4.1. You won’t get any small model like 3.5 Haiku!
It’s mainly routed based on availability, not the query you give it
You cannot use MAX and Auto at the same time.
MAX mode will just unlock the maximum input context a request has, which is where it can quickly have a greater effect on your rate limits. Its not inherently more costly to use MAX, but especially with models that have very large context windows, it has the capacity to be very costly to your usage.
As Auto is of 0 cost, it never uses any of your usage.
Besides Auto, only GPT 4.1 and Gemini Flash have 0 / tiny usage on your rate limits - everything else behaves entirely based on the costs incurred via the models API.
Nope, unfortunately a side effect of the routing system - we don’t know what model will answer your query until it does, and we don’t have a good process for returning and displaying that to you yet!
To additionally answer @qwaker, the window for Opting Out of the new plan has not come to an end, and the option is no longer available to users.
Q: I wasn’t able to opt-out, can I go back to the old pricing?
New users are not able to go back to our old pricing, but for existing users wanting to use request-based pricing until we sunset the plan, you can reach out to our team ([email protected]) and we’ll help smooth this transition.
@FlamesONE@liquefy
If you are on a yearly plan, at least until the end of your billing year. On monthly, at least the end of the month but likely longer than that - a timeframe for when the old plan will be fully sunset is not yet finalized!
@wiireed
Yep, this is still true. To clarify, the self-serve option has been removed from Cursor - The AI Code Editor, but we will still to help with the transition if you contact us via email!
To keep all the questions in one place, I’ll lock this thread for now, but please post any questions in our megathread - the team and I will be checking in often to get back to you all there!
