How much can I use Claude Opus in max mode if I buy the $200 plan?

[kticoder]

How much can I use Claude Opus in max mode if I buy the $200 plan?

[condor]

hi @kticoder, thank you for your post.

The Ultra plan for $200 allows for $400 worth of Ai usage at API pricing.

As Cursor tracks token and model usage, note that Opus is 5x more expensive than Sonnet.

It’s recommended to use Opus when you need an AI that is more advanced than Sonnet. Regular tasks would still work best with Sonnet.

As the actual usage depends on the complexity of project, programming language, framework, prompts, attached files, rules, MCPs,… its not entirely possible to predict how many tokens you would consume.

See more about token usage and how to optimize your usage:

Understanding LLM Token Usage

Here is a detailed explanation of LLM (AI) token usage in Cursor: Tokens are used every time you interact with AI models in Cursor. For transparency, token counts are shown as processed and reported by the AI provider. You can check your usage in ‎Dashboard > Usage and, if you have Usage Based Pricing, also in ‎Dashboard > Billing. The dashboard is being improved to show all AI usage of a single request in one line, even when multiple tool calls are combined. Tokens are divided into four group…

[Alphonse_Jourdain]

Hi,

If you use claude Opus I do advise to pay the same amount of money for claude code, or claude desktop Max tier. You have roughly every 5 hours a reset ability to prompt Opus (with the EUR 100 I get 2 - 3 requests before it hits the maximum every 5 hours, so 1 Opus then switch to sonnet which seems better than Sonnet Max on cursor)

Cursor takes 20% cut on Max models and experience proves for me as enterprise tier, 1 use of Opus burns half of my monthly usage of cursor (500 fast requests, I wasn’t switched to the other thing which is scammy for pro tbh).

I advise to use sonnet on cursor pro for minimal tasks and claude code for the rest. Claude sonnet on cursor has some limitations hardcoded into the instruction set (e.g if you ask to change several files it will state outright ‘okay I’m limited to 3 changes per file, and I have to move on’. Complete tasks while they’re not. Stops at 25 actions so you burn quicker and need to redundantly request the same thing over and over.

That’s my experience feedback.

Happy coding !

[yakovw]

Let me answer you — I have a lot of experience with this…

I’m on the Ultra plan, and to put it briefly: you won’t be using Opus much.
The new way Cursor manages tokens — especially in MAX mode — combined with Opus, eats through your balance very quickly.

And if you ask it to execute a plan, that basically means multiple calls —
because every time you use a tool, it reloads the entire conversation into the model from scratch.
[Happy to explain more if you’d like.]

In short, we’re talking about around 200 calls, each costing you anywhere between $0.50 to $1.00, depending on how much context is fed into the model each time.

Bottom line: don’t use Opus unless you really have to.