Currently, Cursor does not offer any truly unlimited plan like it once did. All plans have usage limits. The current Ultra plan offers $400 in token usage instead of $200, but even that is capped.
Right now, in the near future, users are going to find themselves just a few days after the start of the month unable to use any model!
I fully understand why. To remain sustainable, Cursor must pay providers per token, and anyone who has worked directly with APIs knows the costs involved.
Here is my proposal:
OpenAI is currently the only company offering a truly unlimited monthly subscription. Subscribing directly to OpenAI’s $200 Pro plan gives unlimited access to all agents, including GPT-5 Pro, with no usage limits.
I suggest that OpenAI allow this same unlimited subscription to be purchased through Cursor.
Why I think OpenAI might agree:
They already offer an unlimited plan, meaning they are open to the idea, and they are the only provider doing so.
In their live public demo showcasing GPT-5, it was clear that OpenAI values a strong relationship with Cursor and sees it as a leading platform.
Right now, most developers are still using Anthropic, even after GPT-5 came out, for several reasons — it’s still not perfect, and it’s slower and spends more time “thinking.”
This could be an opportunity for OpenAI to pull developers over, giving them a reason to primarily use OpenAI’s models.
Why Cursor will not lose revenue
Even if such an unlimited plan exists for OpenAI within Cursor, many users will still want to access models from other providers for specific needs. This ensures that Cursor retains opportunities for revenue from additional model usage while providing a premium option for heavy OpenAI users.
I personally would even be willing to pay $300 per month (for example, $100 to Cursor and $200 to OpenAI) for truly unlimited OpenAI usage within Cursor.
lol, when people like @yakovw have cash to burn they stop thinking straight, just throwing money at every problem
I would rather they restore “Manual Edit” mode + true MAX Mode… it is WAY CHEAPER to do ONE request Manual Edit FULL MAX context on a 1M token codebase then MANY subsequent sliding-window-64k-tokens “Agent Edit” mode to correct the remaining simple linter errors, as was the case in previous versions of Cursor
I really don’t understand what solution you’re suggesting. A large context window requires a model that supports it why isn’t agent mode enough for you for that?
I don’t quite get your problem.
And I also don’t understand why Manual Edit specifically solves the problem.
I don’t know if you realize this, but these companies are struggling to make a profit. None of them have made a profit. They’re burning through investor funds. Cursor implemented auto mode because most use cases don’t need the largest, most capable models. Unlimited usage for a flat rate fee isn’t feasible. OpenAI is headed in the same direction as Cursor, but instead of calling it “auto” they call it “GPT-5”. It’s all going to end up being token based usage. Sam has already said he thinks compute is going to be the currency of the future. (That was the hint)
You’re right, but in fact right now Sam’s policy, and OpenAI’s, is to offer such a subscription even though no other company, not Google, not Anthropic, not X.AI, offers unlimited. That means they have some reasoning behind it, so I think they will agree.
You have to understand that they want developers to move over to them, and if there’s something they can offer that will get developers to switch from Anthropic, it’s worth it for them.
