Right now, Cursor offers 500 fast requests — but I burn through those in about 2 days. Rather than limiting the number of requests, it would make more sense to offer unlimited requests, and differentiate plans by response speed.
Suggested Model:
Premium Plan → Unlimited requests, always replies in 10–20 seconds
Pro Plan → Unlimited requests, replies in 30–90 seconds
This keeps power users productive without worrying about quotas, while still incentivizing upgrades based on speed.
It doesn’t work this way. They do not own the model; they still pay for slow requests to model providers. So even slow requests cost them money. And in case of o3 you can easily make them lose a few dozen dollars a month on each user.
So there is no way they will ever remove limits. They are a for-profit company that is actively raising money and has to show profitability to investors.
What if you can’t afford to buy more requests? Wait a year—maybe half a year, until you can use free models locally. They will be at the level of current state-of-the-art models. But top models will always be expensive, at least in the next few years.