Opus 4.7 seems to be more like natural intelligence than artificial …
I am refactoring like a madman and upgrading a large amount of code while the half price sale lasts …
Quite frankly 4.7 is what I had always expected AI to be - finally a model has lived up to my expectations
@KulakovPA Model availability can vary by region, which is likely what you’re running into!
i just got hit this morning with an OPUS-4.7 re-seed cache write on a thread that was active all day yesterday. That cost me 76 bucks for a prompt that was meant to confirm a Cursor bug was still present today. Lesson learned the hard way. Thanks @Artemonim for the info/explantion
Can you get the request data via https://cursortokens.vercel.app/ and send screenshots of the details?
How do I make estimated cost show up when it’s still free included in the subscription?
It somehow pulls it out on its own.
Honestly, I don’t understand why the official dashboard is still so poor.
I’m running the same prompts or intentions for all models to check speed an quality, they judge eachother after, but I’m having difficulties with any method of actual cost evaluation.
I am not really sure, how sometimes input is input and sometimes completely merged into cached, even with different chat, or within the same chat, very mysterious consistency.
Claude appears more expensive, even though it should be cheaper.
i have applied for a $76 refund since i was confirming a bug report I made and the workaround I offered one was incorrect - I think I’ll get it
“Quality” for Opus 4.7 is horrid. Literally waste of money tonight. I just saw it stuff the context window completely with zero output, it doesn’t follow instructions, goes off-side with verbosity and patching together unrelated structures and not only breaks GPT’s work, but GPT can actually fix the mess and continue with real progress.
Hi everyone!
The Opus 4.7 launch promotion will officially end at 11:59 PM PST on April 22.
but i am still running into parameter issues when creating a plan with 4.7 and this model needs to be able to create plans mid thread more than any other model - you are not fully production capable yet with 4.7
Could you extend the discount period? It’s so useful!
Yes likewise, extending the promotion until the weekend would be great if possible!
Opus 4.7 is… interesting. I’m a daily user of Opus 4.5 (I elected not to use 4.6 because of its allegedly cold personality) . Dipping my toe into the 4.7 waters, I’ve noticed that it has a tendency to over-plan and over-architect – a relatively simple request for a security audit plan resulted in a 25+ week, 10 phase project. Opus 4.5 looked askance at 4.7’s work and pared it back significantly.
This is also a plea to Cursor to retain Opus 4.5 access. 4.1 was an expensive token hog, but 4.5 has the perfect balance of competence, warmth and token use (in this developer’s humble opinion).
Hi! Opus 4.7 in Cursor no longer has a promotion, but we have a 50% off promotion on the newly launched GPT 5.5. More info here: GPT 5.5 - Out now!
Hi everyone, so far I’m not happy with Opus 4.7.
I have mainly been on Opus 4.6 and I think it’s absolutely great for tackling research-heavy tasks. Tried Opus 4.7 for about 2 days and for the first time in two years using Cursor I had issues with work being lost. I know that I should have had more safeguards in place, but honestly, until now this has never happened. Opus 4.7 went on with such autonomy and speed that I found myself with the following (for clarification, I had not asked for git reset --hard nor rm -rf!):
These incidences happened shortly one after the other. Again, lessons learnt - add more safeguards and always do all git-related stuff yourself (it was late at night and I just wanted to get it done 
) but anyway, I have never had these kind of issues before and I do find it concerning. Especially taking into account the current trend of just starting (sub) agents to do some tasks in parallel without close supervision.
Bottom line - please Cursor, do not remove Opus 4.6 just yet.
