@bobjr Just to clarify, if you use a free model, there is no slow pool and you should always get a response instantly for the free model, like Gemini Flash - see the models page for more: Cursor – Models & Pricing
@andrewsipe @zixuan0830 If you are seeing ‘Generating…’ and nothing about slow requests, this is likely a bug and not a slow request issue! If you confirm you’re not seeing any prompt about slow requests, let me know and I’ll follow up with you.
@lowcodeai Your first request into the slow pool will be a few seconds of wait, not 5-10 minutes. It will take a significant number of requests before you reach the point of waiting minutes, let along 5-10.
As mentioned in my original reply:
The reason why we offer Pro at $20 is because many users cannot afford to spend more than that, and 500 fast requests is enough for them. Usage-based pricing is the fairest solution here, as you pay for exactly what you use and nothing more. We could easily offer a 1000 fast requests plan for $40, but if you only use 700 of them, you’d pay for a bunch of requests you never use.
@random This system has been in place from the start. We could choose to not offer slow requests at all (as many alternatives do) as 500 fast requests would already cost you >$20 if you did them directly via the API of the provider. We have this feature to ensure people who go just over their usage can still access the premium models.
Think of it like a phone plan. You buy 500 texts for $20/m. If you use the 500, the network provider doesn’t give you any more. You’d have to pay for them, because each text has a direct cost to the network provider.
Giving our more than the 500 means we can’t keep the price at $20/m. Those would are happy to pay more can pay-as-you-go, but those who can’t afford or don’t want to spend more than $20/m don’t have to.
There’s no better way to do this while remaining financially viable, and still allowing the most amount of users to get Cursor, regardless of how much they can pay.