Cursor is teasing a 1.5-trillion-parameter model trained from scratch on over 100,000 GPUs and framed as generally intelligent rather than coding-only, with a launch expected in weeks, yet no official confirmation or technical details have surfaced to back the scale or ownership claims.
New cursor model being teased at compile
- same size as Claude opus and gpt 5.5 - trained from scratch, no more kimi base - 10-20x more compute vs composer - generally intelligent, not just coding
Releases in the “Next couple of weeks”
Public posts mix Cursor branding with details that match xAI’s recent Grok announcements instead, leaving the exact origin of the 1.5T model and its training data unresolved.
If the model materializes under Cursor, developers may gain a new general-purpose agent inside the editor, though pricing, access tiers, and benchmark gains remain unknown for now.
Many users are excited about Cursor's teased 1.5T parameter model trained from scratch on massive GPU clusters because of its ambitious scale and potential, while some users doubt the claims as empty hype or misinformation.
No Digg Deeper questions have been answered for this story yet.
Man it's going to be awesome seeing what the Cursor team armed with SpaceXAI compute can do
Cursor’s last big announcement is a new model! 1.5T+ parameters, retrained 100k+ GPU’s from scratch, intelligent behind coding.
@NickADobos It's Grok v9 @elonmusk was talking about a few weeks back.
@JustinGorya @tibor_tee 🤷♂️ idk
@mntruell said it, I’m quoting. and that they were the same size. I assume this is all an open secret at this point for people in the know
@NickADobos @tibor_tee Where did you get the information that opus (what ever version you named none) and GPT 5.5 are 1.5T models?
I highly doubt that.
@NickADobos 👀 is it gonna be called grok or composer
@NickADobos I wonder if they'll choose a new name for it or if they'll stick with Composer.
@sethsaler I don’t think it will be called composer
@NickADobos Fuck yes, more competition will be good, also Composer 2.5 is already really good
@NickADobos Open AI have fell behind completely. They used to do this and it was Google who would have to watch 5 new models go by before they released their own which would only end up being as good as the previous frontier. 5.6 HAS TO BE BETTER THAN FABLE OR ITS GAME OVER
@NickADobos @AndrewCurran_ I see they've learned the Elon method of promises that mean nothing
@NickADobos woah
@pablonification @NickADobos @sethsaler It's only confirmed when announced
@NickADobos @AndrewCurran_ Why would cursor waste money pretraining?
@NickADobos @sethsaler it is
@NickADobos Composer 2.5 trained on Kimi was impeccable. Hope the standard doesn’t drop. Congratulations
@NickADobos I’m confident it’s just grok with cursors training data, but that’s not a bad thing
@ns123abc Grokky is hulking up.
Grok Build about to MOGG everyone in 2 weeks
@NickADobos @grok так это grok выходит V9, разве нет?
@NickADobos It will be massive
@NickADobos spotted: @jeffbcross's mohawk 🤘