I heard OpenAI is about to cut off GPT access for Cursor — is this true? What would be the impact?
yes we’re losing gpt and this is official. maybe losing Claude next.
Hi all,
OpenAI announced they plan to end their direct partnership with Cursor in three months.
Currently, OpenAI models serve about 5% of total Cursor user traffic. We’re speaking with the OpenAI team to resolve this, and we will share more in the coming weeks around continuity of access to OpenAI models.
We’ll continue to offer ways to access all models in Cursor. Models from Anthropic, Google, Meta, and Grok will all continue to be offered through direct partnership.
OpenAI seems like it is going to be the first of the big companies to fall as they continue to bleed money without maintaining their first-mover advantage. A year ago, this would have been crippling for Cursor. Now? I can’t even tell you the last time I used an OpenAI model within Cursor.
That this was clearly done because Musk and Altman have beef instead of for sane business reasons only demonstrates that Altman is running his company into the ground.
Oh well, grok is quite good already. I have been using it lately because of the speed. When the next version comes out, maybe it will be on-par with gpt sol.
Who are you calling B-tier? Composer 2.5?
Yeah because no other models could possibly exist outside of OpenAI and Anthropic lol. The game is changing, pay attention if you want to keep playing.
Hey @Colin!
I hope the conversation goes well. My team and I really like the OpenAI Models for backend engineering and it would be a shame to see them leave the product.
Hey @Colin
Quick question, I know this is open discussion, but just out of curiosity.
When using OpenAI API, I mean via custom API and custom Model setup, there are big differences in how Cursor treats them.
If for example I create a custom model “my-custom-gpt-5.6-sol”, there is a big difference from Cursor harness between using the “native” Cursor selector for GPT-5.6 Sol versus “my-custom-gpt-5.6-sol”:
- the entire prompt injected differs substatially
- the toolset given to the models differs a lot: “ApplyPatch” vs “StrReplace/Write”, etc
So the question is, in case OpenAI and Cursor partnership ends, would Cursor at least keep the “native” GPT harness, so if we use OpenAI custom URLs, we could still use the more “native” GPT harness compared to the more rudimentary “custom model” harness?
Any news around this?
Thank you!
Guessing now that Scam Altman knew GPT-6 launch was imminent and sure to draw a lot of attention, so cut off Cursor and SpaceX intentionally to try to bring more users to Codex. I’m not falling for it. Honestly, given the compute build out of SpaceX, I’m expecting Grok 5 to be insane.
Hey @Mihai_Cracan
The difference isn’t native vs custom model as such. Cursor decides which prompt and tool set to use by looking at the model name. If a custom model’s name contains one of our built-in model ids (for example anything containing gpt-5.6-sol), it gets the same GPT prompt and the same apply_patch based editing tools as the built-in entry, even when it’s served through your own API key or base URL. If the name doesn’t match a built-in id, it falls back to the generic harness with the StrReplace/Write tools you saw.
That’s why the naming matters more than it looks. A name like custom-5.6-sol won’t match because the check is for the literal gpt-5.6-sol substring, so it gets the generic setup. Renaming it to something like custom-gpt-5.6-sol-medium would put it on the GPT harness. Incidentally, the my-custom-gpt-5.6-sol you mentioned in your post would have matched.
We will share more in the coming weeks around continuity of access to OpenAI models. That news probably won’t be coming from me in a forum thread. 
Stay tuned.
>Currently, OpenAI models serve about 5% of total Cursor user traffic
That maybe true but OpenAI models are very important in our workflows. It is among the best General models, during research/spec stage. Sol and Terra stands out as they are cheaper than Claude models and better than Cursor/xAI models
GPT Luna simply has NO COMPETITION when it comes to how cheap the model is (cheaper than the Chinese and cursor ones) but still performs remarking well
It will be a huge loss if GPT models are taken out. @Colin I hope Cursor team could hash it out with OpenAI so trust can be built again
I am a high frequency GPT model user.
I find them to excel at Unix operating system tasks and configuration, back-end application coding and I still use codex-5.3 for Postgres tunning. Just finished a windows 11 configuration issue with help from GPT-5.4.
And I always play off GPT-5.6 as a planning reviewer against Opus-5 or Fable plans and vice versa. This is my highest value usage for the GPT models.
The loss represents a real blow to my work flow.
Altman is running OpenAI into the ground and this is another foolish mistake on his part.
That’s mostly because they quietly push users off top tier models and if you don’t check almost every prompt, you end up on auto, grok, or composer.
I’m mostly using anthropic models with cursor but lately GPT 6 Astra is something that I’m playing with. And wanna give it a try with cursor but unfortunately it is missing. So guys, do you have any plans to sit and talk with them?
Astra is super legit. It’s the first of the “next generation” models, as far as I can tell. I am incredibly interested in Grok 5, which I expect to be like Fable 5.1 for 1/10th of the cost. And I am also interested in the successor to Fable (odds on what it’ll be called? Folktale? Epic?)
In November, OpenAI will stop providing OpenAI models to Cursor, because Sam and Elon have beef.
This is a shame that is directly impacting the Cursor experience. The main reason I use Cursor is because it gives us great flexibility in choosing whichever model we want to get the job done.
Grok series is great for day to day. OpenAI is great for beefy, token conscious tasks. Claude is great for really beefy tasks where you don’t care about token usage.
The fact that we haven’t been able to try out Astra on Cursor is a taste of this that is hard to swallow. It makes me not want to use Cursor anymore, because Cursor is no longer supplying the flexibility of front tier models.
Is there really nothing that can be done to, maybe not end the beef, but at least put it aside for the Cursor user?
Well, by the time this happens we might have a new Grok version and/or composer. I can imagine xAI is feel ‘more’ competition to develop better models. (lets be honest… 4.7 is way too behind)
I’m skeptical. Grok 4.7 is not near the Opus models it claimed to match, so I doubt 5 will be at Fable either. I stick to Grok because the plan has far far more allowance for Grok use, but I’m considering to cut my Cursor plan even though I really like the Cursor and Grokbot UX.