We've hit a rate limit with the provider. Please switch to the 'auto-select' model, another model, or try again in a few moments

Discussions
1

This alert has been appearing all day today. Do you have any questions?

We’ve hit a rate limit with the provider. Please switch to the ‘auto-select’ model, another model, or try again in a few moments.

2

We’ve hit a rate limit with the provider. Please switch to the 'auto-select

@CURSOR Didn’t you say the PRO version is unlimited?

3

Cursor doesn’t work at all today, and I’ve been using it very infrequently lately.

4

I also have this question! I don’t know why. Cursor need let us know the real limit

1 Like
5

This is not about the user rate limit. If you read the OPs message in detail it says Cursor has reached their rate limit. Likely the AI provider has limited capacity or large number of Cursor users suddenly use the same model.

This is usually temporary.

2 Likes
6

I got these problems from yesterday and till now it is showing the rate limit reached and i am not able to use this whole day today. I don’t know even when it will be reset and how the rate limit reached just after one prompt ‘I just give one simple prompt saying Hello ’ after this prompt when i gave next instruction to do then it gave me rate limit warning yesterday and till now (after 20 hours) still showing same warnings. I can not use any models except auto mode (worst) in auto mode it uses chatgpt 4O and some time Chatgpt 4.1 only. I tried to find official documentation for this rate limit but did not find much information about this. Can any one tell me why these problems occurring?

7

HI @user100 could you share a bit more about your usage?

  • Which model did you use?
  • Which mode (thinking? max?)
  • Did you try switching to antoher model? (e.g. if using Opus which is resource heavy, switching to Sonnet or Gpt-4o would work)
  • Were there any attached files to the chat (e.g. files, folders, rules, …)?

Note that not just heavier models consume faster but also more context, so managing context helps a lot with usage.

Yesterday I used on my personal Pro plan for 3.5h Claude 4 Sonnet Thinking Max on one task without reaching any rate limit. I let Agent fetch any content it needs, and just told it to process tasks from file xxxxx.md (not attached). When it finished I had some follow up requirements.

[Edit: added following questions for better understanding of limits]

  • How many requests were made?
  • How long did it take you to reach the limit?
8

I mainly use claude 4 sonnet (thingking) today, and this warning is displayed most of the time

9

Could you answer the same questions I posted to another user above? The help understand how you reached the limit.

10

Today I mainly deal with some simple bug fixes and some simple functional businesses, without any particularly large context. Sometimes I will attach some pictures of the bug interface.

I did not switch to other models. I usually only use this model because I am not familiar with other models. So when this situation occurs, I just click to retry. Sometimes it can continue, sometimes it will become very slow, and sometimes it can’t continue and still prompts like this.

Because my task is not urgent, I just leave it alone.

1 Like
11

Thanks for the update. I assume that pictures consume a lot of resources as they need to be converted into base64 encoding for data transfer and processing.

Did you ever try an MCP for browsers? that would avoid having to send full screenshot as Agent could access the bug in browser.

12

I used partial screenshots, which should be very small, about tens of kb but less than 100 kb. I have never used the browser mcp.

I will try it later to see if there are any problems. If there are any problems, I will continue to report them.

1 Like
13

I was using gemini 2.5 pro (not in max) and when i got this limit warning I switched to Claude 4 sonnet (thinking) and the. I do one request and after this i got the rate limit again and then i was not able to use any gemini model or any claude model. But when I switched to auto mode then it used chatgpt 4O which is worst for complex tasks. I used gemini approx 2-3 hours and then i got these limit warning and after that i was not able to use any model. Till now its has been 20 hours still showing the warnings. Do you have any idea what happening and how this rate limit system work actually?

14

No I have no direct insight how the rate limits actually work, thats why I tried also by myself to see if I can reach a rate limit. I see several posts in the forum where users have used it without being limited yet, and a few others have run fast into limits.

As the models you used are not the heavy models (o3, Opus,..) I think that some of that may be due to context size of chat. (Any large files attached or read by agent?)

I think that such cases like yours would need more details to understand what is happening since regular usage should not reach the limit so easily.

Have a look at my posts in following thread. It shows my yesterdays work on Claude 4 Sonnet Thinking Max with personal Pro account. (3.5h work non stop without reaching rate limit)

1 Like
15

Hey just before 10 min i got the access of Claude 4 Sonnet Thinking (not max) so I tried to use and test it i just gave it small tasks and it worked but after two prompts it gave me the limit warning again. I wait for about 20 hours and it gave me same limit warning just after two small prompts. Why??

1 Like
16

Its nearly impossible to say without knowing more what you tried in detail.

I would say it could do with how much context was added to chat but thats just a guess as you would have to share a bit more details.

Do you have a request ID (preferrably with privacy disabled) for the Cursor Team to look into?

17

Of course this is my last chat or request id which i did with sonnet 4 thinking (not max). e414aafb-53c0-45b6-8216-1ce426120035