I have a few questions and concerns about Cursor’s model roadmap, and I’d love to hear your thoughts.
1. Grok 4.7
When will the Grok 4.7 model be released?
2. Composer 2.5 → 2.6 / 3.0
It’s been several months since Composer 2.5 was released, and there haven’t been any updates since. This doesn’t match the pace of a normal model iteration cycle at all. I’d like to know:
When will Composer 2.6 (or Composer 3.0) be released?
Or is there no further update planned for the Composer line?
3. Cursor’s model strategy
Is Cursor planning to shift its full focus to the Grok models? I’m honestly not sure how Composer and Grok are positioned within Cursor:
Are they competing with each other, or complementing each other?
4. Model competitiveness
I also feel that both Composer and Grok are now falling behind third-party models (such as the latest Claude and GPT models).
I still believe in Cursor and I’m still choosing to use it — but my confidence isn’t as strong as it used to be, and that makes me a bit worried.
Would appreciate any official clarification or community insight. Thanks!
I have an an annual plan with them (with another 6 months to go) and am starting to feel buyer’s remorse at this point.
Neither are the models catching up nor are they’re ready to provide a pro-rated refund.
Its like being stuck between a rock and hard place.
Update:
I just realized that I was not providing a wholesome perspective in my comment above. I think I would be remiss if I did not mention the amount of value that I’ve gotten from Cursor since I took the annual plan towards the end of March. It’s completely changed the way I work and the level of productivity that I’ve been able to have. I’m very, very thankful for that.
One of the things that bothered me, though, was the fact that I had to invest more money for the Frontier models, even in spite of the fact that I already had the annual plan (which gives about $400 worth of tokens for the non-Cursor models). I think I’ve invested almost $300 extra in the last 6 months that I did not anticipate, and I might not have had to do that if Cursor and Grok were actually up to the mark.
Having said that, once again, I think the product is fantastic. I think it works really, really well. I still haven’t used many of the features in Cursor, but with whatever I have used, our product has just come such a long way, much further than I would have ever expected.
I might have gotten the same thing from Claude Code or Codex as well, but considering I am in Cursor and at least that part of the promise was lived up to, I think it is still extraordinary.
There’s been lots of discussions about this in the Composer 3 request thread as well. Based on communication from earlier this year, I believe the original Composer 3 work went into Grok 4.5, but we definitely need a replacement for Composer 2.5 - and soon. GLM-5.3 Flash, Deepseek v4.1 Flash, and even Deepseek v4 Flash 0731 are all far more intelligent models at a similar (or even less) cost-per-task that diminish the value proposition of Cursor day-by-day. Grok 4.7 is also necessary to compete at the high end especially with the OpenAI departure in November.
A confirmation on if we’ll be getting a Composer 3 (i.e. an RL’d open-weight model) or a Grok “Lite” (i.e. smaller Grok model) would also be appreciated, as you’ve laid out. They mentioned in the Composer 3 request thread that they’ll “continue to release models of this size”, but everyone kind of already knew that.
Hey, thanks for the detailed and thoughtful post. Point by point:
Release timing for Grok 4.7 and Composer 2.6/3.0. We don’t announce exact release dates or unreleased models in advance, so I can’t share a specific timeline. All new models and updates show up in What's New in Cursor — Latest Updates & Release Notes, that’s the most reliable place to follow.
Composer vs Grok, our strategy. The Composer line isn’t closed, and we’re still working on models in that class. Grok and Composer serve different goals, frontier-level performance vs intelligence-per-dollar, so it’s not an either or choice. I can’t share more detailed positioning until there’s an official announcement.
Competitiveness. The feedback that Composer and Grok are behind the latest Claude and GPT models is a valuable signal for prioritization.
Some of the discussion is already happening here: Please release Composer 3 soon. I’d recommend subscribing there if you want to keep up with the topic.
Separately for @Vaidy_Bala, questions about the annual plan and refunds are handled by the team at [email protected]. Send them a message and they’ll look into your specific case.
Remember the main selling point of cursor is generous allowance of usage with Composer and overall other models. If we feel Composer 2.5 is not catching up enough, the grass would seem greener on the other side. I am on monthly plan, so ready to make the switch anytime.