Opus 4.7 context window (increase non-max above 200k)

Ideas Feedback
, , ,
1

Opus use to charge twice when the context window went over 200k, cursor was great by letting us restrict it so it doesn’t go over.

But now Claude does not charge extra for high context usage anymore, but cursor still has the non-max opus context window at 200k. while chatGPT gets a non-max context window of 272k, which doesn’t seem fair (cursor charges their own extra fee when max mode is on).

Also Claude updated how they tokenize context so now it uses more tokens for the same prompt. Every time I would fix something with opus it would use 170k tokens, then I clear the context for the next fix. Now each fix uses 210k tokens which means I always have to have max mode turned on, because the context is restricted so much more then chatGPT is for no reason now. So now I’m paying the unfairly paying the extra cursor max fee. please increase the standard opus context length to match chatGPT

3

If you’re on the most recent update of Cursor, you can click on the model selection dropdown, hover over the model you want to use, click the Edit button that appears on the right, and then you have options to configure the context, effort, and toggle thinking.

image
image789×465 18.1 KB

This behavior is new and not very intuitive. I also had trouble finding it today when I wanted to change the effort selection:

Note, that I haven’t tested this yet to see if it actually works, but at least it looks like the options are all there.

I hope this is what you’re looking for and can help you too.

Edit: It looks like selecting 1 M Context automatically toggles on MAX Mode as well, so this doesn’t help with what you were asking.

5

Hey there @zebra !

Good news – for users on individual token-based plans, there is no longer a 20% markup for using Max Mode.

Hope this helps!

7

Is this officially stated anywhere? I cant find where this is announced @Colin

9

Hey @zebra

We’ve updated the docs, but no further announcements planned!

On current individual plans, Max Mode is billed at the model’s API rate. On legacy request-based plans, Max Mode adds a 20% surcharge.

10

Thank you!

11

@Colin
That clarification honestly makes this feel worse for legacy users, not better.

So current individual token-based plans get Max Mode billed at the model’s API rate with no markup, but legacy request-based plans still get hit with a 20% surcharge? That feels like early users are being punished for staying on the older request-based pricing instead of moving to the newer token-based system.

The bigger issue is that higher context now seems increasingly tied to Max Mode, which means legacy users are effectively pushed into a more expensive path just to access the same practical workflow. 8 requests for around 120k tokens, Max Mode only, plus a surcharge, feels unreasonable.

Cursor already lost a lot of trust with the previous pricing changes. Continuing to make legacy plans worse while quietly updating the docs instead of clearly announcing the change feels like another blindsiding of the user base.

14

Actually can you clarify that. “legacy request-based plans” is weird wording. I have a annual plan (non-commercial). do I have to cancel my plan, and then make a new plan? Or is this dropping the 20% is for everyone right now? (also I noticed that non-max context has been increased to 300k, thank you).

15

@zebra You’re on a token-based plan. Your usage is measured by the number of tokens consumed, not by a fixed number of requests.