Opus 4.7 - Out Now!

Announcement · Pricing


Claude Opus 4.7 is now available in Cursor!

Opus 4.7 is now available in Cursor. We’ve found it to be impressively autonomous and more creative in its reasoning.

Have you tried it yet?

We’re launching it with 50% off for a limited time. Enjoy!

7 Likes

Is it only the ‘max’ mode? Is this done on purpose to drive away legacy users?

12 Likes

Please allow us to toggle between API pricing and the 500 request pool so legacy people are able to try the new models as well.

I’m not asking to use the new models in the request-based system. I’m literally just asking to be able to pay for it using API pricing.

4 Likes

Is it available for Cloud agents yet?

When will it available in non-max mode for legacy user?

1 Like

(post deleted by author)

I’m running into some issues with it. It seems to continually hang on the first api request. Here’s the
Request ID: 178ba164-6745-4e7c-adc1-bf3157ceddb4

Error Message:
[internal] internal error
internal error
ConnectError: [internal] internal error
at t (vscode-file://vscode-app/Applications/Cursor.app/Contents/Resources/app/out/vs/workbench/workbench.desktop.main.js:38575:65674)
at async Object.addAsyncFollowupBackgroundComposer (vscode-file://vscode-app/Applications/Cursor.app/Contents/Resources/app/out/vs/workbench/workbench.desktop.main.js:29031:4386)
at async fzd._sendFollowupRequest (vscode-file://vscode-app/Applications/Cursor.app/Contents/Resources/app/out/vs/workbench/workbench.desktop.main.js:45160:30894)
at async fzd._submitMessage (vscode-file://vscode-app/Applications/Cursor.app/Contents/Resources/app/out/vs/workbench/workbench.desktop.main.js:45160:29572)
at async fzd._submitMessageForAgent (vscode-file://vscode-app/Applications/Cursor.app/Contents/Resources/app/out/vs/workbench/workbench.desktop.main.js:45159:68565)
at async pxC.submitMessage (vscode-file://vscode-app/Applications/Cursor.app/Contents/Resources/app/out/vs/workbench/workbench.desktop.main.js:45159:3772)
at async vscode-file://vscode-app/Applications/Cursor.app/Contents/Resources/app/out/vs/workbench/workbench.desktop.main.js:43244:89272

@Mikil_Foss Are you trying to use it with a Cloud Agent? We’re working on getting that working right now (@Nick_Loeper)

Ah, yes I was. FYI, I am getting it working on a cloud agent thread; I started the thread on a different model and changed it on the desktop app to opus 4.7 and its working well.

https://cursor.com/agents/bc-25f0da14-9b32-4b8b-b53b-ecc133ec57b2

Hey everyone!

This thread is for discussing Opus 4.7: how it performs, what you’re building with it, first impressions, etc.

Opus 4.7 follows the same pricing policy as GPT 5.4 and other frontier models going forward. If you’re on a legacy request-based plan, you’ll need to enable Max Mode to use it.

I get that there are strong feelings about recent pricing changes, but we aren’t going to let this thread devolve into a new debate about this policy. Off-topic posts that violate our community guidelines will be removed.

1 Like

Can I ask how long the limited time period is for the 50% off?

2 Likes

It’s good, but even with a discount, it’s expensive. When Mythos Fast comes out, it will be an AI for Arab sheikhs :smiling_face_with_tear:

how do you get this info?

Opus 4.7 experience so far:

  • never runs any tools even if I tell it to
  • keeps running all kinds of commands, mainly ls, cat etc to read folder structure and files
  • even on max effort it does not really think for long, it started running commands instantly
    lol garbage model

For me, it only did this for the directory in .gitignore.

And it seems to work well with subagents.
The report has been translated into English using Qwen3.5.

but expensive :smiling_face_with_tear:

Using it for 2 hours now, especially with the 50% off (the price drop does not appear on every session?). It works fine for know, especially with depth on coding without forgetting things. But it uses an abnormal amount of tokens. I have never used this much tokens in one try since a year…

Perfect timing. I have a fairly large architectural redesign to do today and a token budget of $100.

1 Like

How long is the launch window @Colin ?

Do you not treat legacy users like human beings? Just two questions wiped out all my requests

1 Like