On the model page (Cursor – Models & Pricing), there is no pricing information displayed to the user. However, looking at the page source code, the pricing can be determined. I learned this by querying an AI and then I checked myself. It seems rather important since the cost of one request of GPT 4.5 or Sonnet 4.0 will impact how fast I get throttled. It is really disappointing that you (Cursor) do not provide more usage information to the user. You ask the users to trust you. Trust has to be earned, especially these days.
Hey, with the new pricing model, the details about per-model pricing no longer apply, which is why they have been removed from the documentation.
What you are likely referring to is some leftover code with these old model prices that can still be found in the HTML - I’ll make sure to get this removed soon.
We still need more information about pricing in the following cases:
Some users may want to opt out of the new pricing.
What happens with Max Mode?
What about usage-based pricing?
Also, according to Reddit, there are serious bugs when switching back to the 500-request Pro plan. Now, every tool call is counted as a request. And I just confirmed it. Although the dashboard’s usage section doesn’t show the request count, if you check with Chrome DevTools, you can see that the internal API reports a much higher number of requests than before.
I think you misunderstood my point. The three points I mentioned are not questions themselves—they are reasons why pricing information is still needed, since this info has now been removed from the documentation.
Also, I’m aware that we can opt out on the dashboard. However, has the bug been fixed? I’m worried that if I opt out, my request count will increase by 100+ because every tool call is counted as a request.
Or will the rate limit hit for ALL models/usage? And is there a timer currently for when the rate limits reset?
I really dislike the rate limiting approach personally. When I’m doing Cursor, it’s basically a very few concentrated days of the month, where I burn through like 50-100 on a day, usually in a single “session” (over multiple hours, depending on the task). Then I’ll probably have a week or so with no requests made, and then later in month another day with a 50-100 requests session.
I feel like the rate limiting is more for users who’re using it daily, but in smaller bursts.
This is a bug, as the requests system is no longer reporting in the same way as the old one, so the usage here does not reference the old usage system!
Question about new Pro plan prompt limits (compared to old model)
Under the old pricing model, I was using some pretty complex prompts—think: 15+ tool calls, 10+ searches, and lots of tokens. Each of these would count as roughly 0.8 of a Sonnet 4 “request”, so I could run over 500 of them before hitting the limit and getting moved to usage-based pricing.
With the new Pro subscription model, does anyone know how many of these kinds of complex prompts I can now run before I hit a cap and getting moved to UB pricing?
Trying to figure out how the new limits compare in real-world usage.
Hey Dan,
Thanks for the info For the background agents.
To clarify, Max is available in agent mode, and for background agents it’s usage cost only? Cursor Pro & Cursor Ultra - #38 by chezluc
Thanks!
Thanks. I just opted out, it worked as expected. But there is no request count stats in the usage and the dashboard homepage, hopefully these will be restored too. Thanks.
Does the lack of public transparency mean you can control any model’s traffic at will, without providing us with any information? This is no longer a matter of security; it’s a question of consumer rights. We hope you understand that from a consumer’s perspective, we need a complete understanding of how this Rate Limiting operates and what the conditions are. Instead of passing the buck or being evasive, we urge you to be clear about this. Your continued refusal to clarify this matter is akin to breaking the law. Digital products that do not provide clearer information to consumers about their content directly impact consumer rights.
For the background agents.