Hi. I was considering re-subscribing to Cursor; I haven’t used it since before Claude 4 was released.
On the pricing page, it does not list Claude 4 Sonnet and Opus as premium models. I don’t know whether this is a mistake.
I’ve found similar posts but couldn’t find a conclusive answer. Do you have 500 fast requests and unlimited slow requests for 4 Sonnet and 4 Opus – or only 500 fast requests? If you run out, do you need to pay for Claude 4?
Apparently, 4 Opus only works in Max mode, meaning it’s pay-as-you-go on top of the $20 a month. What about 4 Sonnet?
Hey, yes, Claude 4 Sonnet and Claude 4 Opus are available with a Pro subscription. Claude 4 Sonnet can be used in normal mode, while Claude 4 Opus works only in Max mode. In this mode, you can still use your fast requests from your subscription, and pricing is calculated based on tokens.
It’s important to note that both Claude 4 models are not yet part of the slow request pool. This means that after exhausting the monthly limit of 500 fast requests, further use of these models is only possible with usage-based pricing enabled.
More detailed pricing information is available here:
I don’t understand. The LLM model selector popup says that Sonnet 4 Thinking uses 0.75x and Sonnet 4 0.50x requests. Why would cheaper models (that use less than < 1 request) not be in the slow pool? Where is that criteria documented? Could you please clarify that?
I’m very worried that your screenshot and information might help them come up with a very effective solution — changing 0.5x requests to 1, and 0.75x requests to 1.5.
Slow pull became unusable.
I usually tend to use like 600 requests a month but dude with the current state of slow pull I can just use a local model without any subscriptions so the slow pull is basically void as an option. And let me guess, the slow pull became so bad because of all those free student licenses? Maybe you had to divide paying customers and free licenses in the slow pull to make it work, Cursor?
