Your link doesn't suggest a violation of TOS, but rather a policy shift disallowing their major competitors to use their model for any use.
That thread you linked suggested they were using it with cursor as part of their development process, not for distilling.
>"Hi team, I believe many of you have already discovered that anthropic models are not responding on cursor. According to cursor this is a new policy anthropic is enforcing for all its major competitors."
what's the moat for AI labs like OpenAI and Antrhopic as they seek trillion IPOs. a lot of people said Chinese models also distill US LLM models. if it is this easy to do. how do US AI labs justify asking for trillion?
I believe if this was true, we'd already be seeing vibed stuff succeeding everywhere, and pricing out incumbents, but I guess I'm not seeing this? Like at all?
Even if they have Astra or whatever completely to themselves, how do they deal with the -product- side? Or sales, or account management?
They've been focused on juicing model's coding capabilities, but it's absolutely -not- "gen ai" enough to be doing the whole thing in agents, even 3 or 5 years from now, if only because there's so much context we _can't_ give to these models, without reverse centauring ourselves with cameras and mics and oodles of compute everywhere, and so far that hasn't exactly been playing out the way the frontier labs and the singularity folk hoped it would (meta glasses? humane? really?)
Anthropic wants you to think that models that distill from them would be worthless otherwise. It’s not true, it’s just one part of the process. The whole narrative that Chinese and other models are only good because they distill is nonsense.
Oracle makes billions selling SQL when you can just use a free version. This is no different. Enterprise has enterprise needs. It's really not that complicated or irrational.
i'll be surprised if anthropic doesnt follow. cursors biggest advantage has been giving users a polished wrapper around whichever frontier model they prefer
Not being able to read about Musk doing some shit because Musk have done some shit to Nitter is just peak comedy.
Just go ahead and do set it up. You’ll be my hero (I don’t have mental capacity to research which jurisdiction works for this right now, but obviously not in US or EU.)
LOL, obviously Space-cursor-xAI-musk-imperial-holdings-inc is providing them compute. They either _can't_ cut them off, or they know not to bite the hand that feeds them.
Yes Anthropic’s chief compute officer said nice things about one of Anthropic’s chief providers of compute. Recall that compute constraints are one of their existential risks.
Lol, Anthropic runs a significant portion of their compute on xAI hardware, wtf are you even talking about trying to suggest this nefarious shit?
Some of you guys hate Musk so much that you say blatantly nonsensical stuff. Like, how can you be on HN and not have some idea about the xAI-Anthropic compute contract, or the fact that xAI took down Grok capabity so they could sell compute to Anthropic? And if Anthropic hated Musk/xAI as much as you are saying they do, why would they be paying xAI billions for compute? Think before typing.
It's ironic how AI companies are re-inventing their own form of privately enforced copyright, lobby the government to ban foreign competitors that don't respect it etc., all while spending the last 5 years fighting tooth and nail against the copyright of the training material they're using.
If you can take any book and turn it into a model, because it's "transformative enough", and "AI learns just like a person does", then surely a model distilling another model is transformative and fair use.
They tied themselves into knots fighting the letter of the law, and now, when they need protection in the spirit of the law - that each creator deserves protection for their work - now we devolve to the law of the jungle. Maybe we'll even see LLM book curses.
I may be a rare Cursor user around here, but I am saddened by this news. The thing that keeps me using it is precisely the ability to switch between models: I usually do brainstorm/planning with Sol or Terra, then let Composer (free) deal with the implementation. For certain tasks I use Claude, which tends to be more expensive but sometimes saves me time.
Using both OpenAI and Anthropic models, Cursor is cheaper than having two separate subscriptions, and their free tier of Composer is good enough. I will now have to reconsider my options.
If you can explain how to make compaction not suck in opencode or Kilo Code, I would love to know how. I am really trying to make use of other harnesses but that's a major sticking point.
Never compact. Not automatically, not manually. Instead, let it write ledgers and checklists or handover prompt if necessary, and start fresh sessions as often as possible.
currenly I use pi with openrouter and is very nice I combine with a chatgpt sub that allows me to use it with pi. and when my usage runs out I turn to openrouter. Then same when I need anthropic models
JetBrains is a good option - works with any model/agent. Also they have very good semantic search for the codebase, which is exposed via MCP to the agents.
It's always been clear to me that Cursor's business model of reselling others' APIs had its days numbered. Not necessarily because the providers would pull the plug, but because you wouldn't be able to compete with subsidized plans.
Cursor is already kind of useless for third party models unless you're willing to spend thousands of dollars. It's only worth it if you're going to use mostly grok/composer.
What do you mean by that? If you're being sassy about a downvote then I'll remind you that when you reply to a comment, that person can't downvote your reply. Also this is an account with barely any comments or points, I'm pretty sure they can't downvote at all.
On one hand, yes, it doesn't make sense to give out subsidized tokens forever.
On the other hand, subscriptions create lock-in in a way that API pricing doesn't.
I think a more likely end is that subscription value decreases over time because API pricing gets more reasonable, but subscriptions stay because they are a good way of getting money out of people consistently.
Right, subscriptions are not only cheaper because they "create lock-in". They also let you do things like forecast demand and plan your capacity for it, borrow against it sometimes if you need to, and keep somewhat less around because your cashflows are predictable. It is also valuable because it keeps customers around but that's by no means the only reason recurring revenue gets a higher multiple.
This is why "subsidized tokens" is possibly a misnomer. Money at lower variance is worth more than the same money at higher variance. Not "subsidy" so much as reducing risk and passing some of that to a consumer.
If the Opus/Fable harness is superior and it's written using Opus/Fable then it will cease to be a moat. Otherwise the premise for coding using LLMs is essentially untrue.
I've recently moved to a company that uses Cursor, and I'm actually quite fond of it.
I basically never use the editor but the fact that there's a review UI for all the agent work is incredibly helpful.
For me, that makes the models much more usable. I've also been using GPT models a lot, as they're cheaper and less vomit inducing than Claudes text, so this is definitely bad news for me.
I think the point is that CC was the first general use, terminal based harness. I think. Now you can use PI, or OpenCode, or whatever, and use any model you like.
Now, I’m no expert, because I was late to the game and have only ever used Pi. But I guess Cursor is some product that’s tied to your IDE? If so, then yeah. That’s just too restrictive. I still love my IDE, but I don’t want it to be my harness too.
> Claude code is a decent harness but then you have to use Anthropic models...
Claude Code works with models from other providers too. Anthropic supports this. You can configure some Claude Code environment variables to switch: eg changing ANTHROPIC_DEFAULT_HAIKU_MODEL to point to GLM Flash or Luna, setting ANTHROPIC_BASE_URL to point to api.z.ai, and making ANTHROPIC_AUTH_TOKEN the API key for your alternative provider instead.
That said, I've not actually tried this myself, opting to build my own harness instead. And you don't know what information Claude Code might be sending back to Anthropic about how you use competing models and which models you use. I don't know for sure that they do this, but after hearing about how they used steganography in the date of harness system prompts to identify the user's location, I don't entirely trust Claude Code anymore.
I saw that happen culturally. But never saw that much different in performance. And I really dislike terminal development. And their desktop app is really buggy.
The tide is turning though. I was using the old non agent view in cursor every day for about a year when I built my startup with it. Then Claude Code blew it out of the water. But since two month or so I much prefer Codex and Cursor with the agent view.
However, what I see cooking at Cursor is way more promising than the others. Things like full system understanding, their own forge, multi-repo support. I'd put them as way more visionary in the Gartner magic Quadrant ;)
Why do people keep insisting that subscription plans are subsidized? They are not. API prices are outrageous, designed to milk enterprise users. They probably have an 80-90% profit margin on subscription plans.
Inference is cheap; Dax once said on a podcast that Opencode has a close to 90% profit margin on the openweight models it provides inference for. Those models are close in size and spec to models from large labs.
What a shame for Cursor. I found it a great tool - I actually liked Composer 2.5; but now being bought by SpaceX has for a while left a bad taste - and they keep pushing Grok and now GrokBot.
I took the 50% discount instead of cancelling - but with this I'll likely not renew.
So sad to watch them go from great to mediocre in such a short time.
While this isn't great for users it seems like a pretty standard circling of the wagons in preparation for the next phase in the battle for frontier AI dominance.
Ironically, as a Cursor and Claude subscriber, but not an OpenAI subscriber, this will push me back to Anthropic. I don't feel good about giving OpenAI money directly, but I've found switching to GPT 5.6 Sol now and again in Cursor to be useful. Now I'll probably just forget about OpenAI models.
Exactly. Astra will likely get its Mythos moment, and then another Chinese lab releases its equivalent three months later while Anthropic releases a newer, better model.
There's a reason why most businesses don't hire PhDs, they dont need it. PhDs are cheap in many countries, arguably more affordable than a full time mythos lvl agent, yet businesses in India still don't hire 30k a year PhDs to run their businesses. Its simply not necessary. AI labs have lost the plot from a business perspective. If chasing ASI is the goal, that also seems stupid.. Putting all of your customers out of business with ASI doesn't really lead to great outcomes either.
Yeah I haven’t had a problem yet which was too difficult for Fable to solve. I do have problems which Sol does not do well on. But if I could get Fable-level intelligence faster/cheaper and less restricted, that’s all I need.
Fable is great minus it's annoying writing style. It's nowhere near as bad as Opus 5 but it's very "AI trope-y" even with heavy steering. I can spot the writing immediately.
Yeah not sure why everyone thinks they need a PhD lvl assitant to build their slop SaaS when an open source model that is 100x cheaper, and 3x faster and arguably still "PhD" level assitant could also build their slop SaaS.
> as a Cursor and Claude subscriber, but not an OpenAI subscriber, this will push me back to Anthropic
Instead of deciding based on the merits of the model, they are being pushed by the actions of the companies. You should be able to use the best model, or whatever model you like, independently of what any company thinks or does.
It’s OpenAI (and Anthropic I’m sure) realising that they’ve somehow ended up stuck in the lowest margin part of the stack (providing inference), and trying to get out of it
I don't see how it makes sense to use models outside of Grok or Composer in cursor unless you like the tooling to such degree that you're willing to pay out the nose for access to OpenAI and Anthropic to use with that tooling. And Codex is regularly scoring towards the top among harnesses, so you might as well use the best harness available for the given model. I'm happy with Grok and Composer in Cursor. If Cursor wants to add more models, they should host more open models.
I can’t speak to the veracity of the benchmarks but it appears their methodology is sound. Nanobot has 47k stars, fwiw https://github.com/HKUDS/nanobot
It has been more of an OpenClaw or Hermes alternative than a coding agent like OpenCode or Pi, so it’s likely to do well given less context bloat.
The reason it makes sense is because different models are better at different things. Allowing yourself to be locked in to a particular model by these companies is a bad idea and will lead to worse outcomes for yourself and the market as a whole.
If you have such a world changing technology, why sell an API when you can capture all the value yourself? That's what every provider is asking themselves.
Cursor will be fine after this, as they have their own first party models and can potentially use open weight models via API or their own hardware, which have gotten good enough to Opus and Fable quality now for many use cases.
Being a pure play model provider is completely untenable right now. Coding agents own your demand; Nvidia, data centers, and DRAM manufacturers own your supply. You're getting squeezed by open weight models. Your only bet to making money off this thing is by locking your model down to your program to have some semblance of a moat. The fact that businesses can change providers with just the change of an API key is unsustainable. You have neither the "hard to switch" lock-in of a traditional software company nor do you have the near-zero marginal cost of production they have. Each token of inference costs you money, not to mention the training costs. At least locking the models down to your harness lets you get some sort of lock-in.
It makes no sense that a contract would even provide a maximum notice period. Why would parties put that in? What benefit would either derive from restricting themselves to not getting earlier notice.
I read it as: we acted as quickly as we could once it became possible to do so, i.e., the change in control was completed. not that their special agreement carved out some maximum notice term.
Everyone is playing hardball. No one wants to be IBM in an IBM-Microsoft dance.
But this is kind of like banning GPU exports to China. It will encourage them to develop their own technology. Yes, it will take them longer, but in the long-run China will be more successful with their own GPU tech.
> As AI capabilities advance, we also have a new level of accountability to ensure our upcoming model, Astra, is being used in accordance with our terms.
I think Sam Altman has played a clever business move here, they can use this excuse immediately after the Hugging Face leaks and it's not going to be dismissed willy-nilly. Whereas long term they probably wanted to end their partnership with Cursor anyway. I bet Dario will be happy about this too, but maybe the compute deal prevents him from doing similar anytime soon.
Is it really an excuse? As the release points out, musk admitted under oath to distilling OpenAI’s models to produce xai models on an ongoing basis. Whatever you think about goose for the gander wrt IP theft, I wouldn’t voluntarily do business with a competitor who was actively and brazenly flouting my terms, would you?
> As AI capabilities advance, we also have a new level of accountability to ensure our upcoming model, Astra, is being used in accordance with our terms.
This move seems to protect Astra against distillation. I wonder whether Astra will be released before or after 11/12.
My guess is that OpenAI now has a full blown Cursor equivalent in development (an expansion of Codex?) and will announce it in September. Otherwise it stands to lose those who are used to working with Cursor.
Yeah, the "go above and beyond for our users" implies they have a "better" version of Cursor coming. Probably integrating their new chip/microphone based product.
Honestly, can't say I miss my Cursor subscription at all, and I'm surprised they're still a going concern. Why would I want to use a proprietary VSCode fork with an identity crisis when I can use literally any editor with any agent and have about the same experience?
> We know that the people most affected by this decision are the developers who rely on OpenAI models in Cursor. We care about their experience in this transition and we’re ready to go above and beyond to support them.
How so? A dozen very capable models out there. And the gap between frontier models from the two main US labs kept narrowing, at this point non existent.
The quality/price ratio is clearly superior when running Qwen, GLM, or Deepseek.
To make a product more exclusive won't hide the fact that product is becoming more and more irrelevant.
It was honestly just a matter of time, given all of the bad blood there is between Sam and Elon and Elon basically admitting under oath they were distilling other companies models.
I wonder if Anthropic will follow suit given their campaign against distillation or if they'll stay quiet given their reliance on SpaceX (among others) for hardware.
Not surprised, but honestly I don't feel like this is a huge loss for cursor. Models are so commoditized, and 99% of users don't need particularly fancy models. Between existing anthropic models, composer-2, honestly even grok models, and also the fact that many people just use 'auto' mode, I feel like many won't feel the impact for OpenAI's withdrawal.
Maybe if Claude joined along it would be more impactful. For better or worse though, Xai seems like its lining themselves up to handle being on their own for cursor.
At my dayjob we are big users of Cursor's Cloud Agents. We use them to trigger automated investigations customer support issues, alarms, feature requests etc. The Cursor editor is fairly meh in comparison, but this has become a big part of our workflow. And having anyone on our team manage them. What competing products are there we should try?
Elon Musk openly admitted to distilling OpenAI's models [1], which I think explains the "based on our experience with Elon Musk's companies violating contracts".
Last year, Anthropic cut off Windsurf the same way [2]. Fortunately, unlike Anthropic, OpenAI allows you to use your subscription with other harnesses, including Cursor.
Of course today, Anthropic is taking the "high road" and is OK with with SpaceX [3] including their models in Cursor, because they need SpaceX's compute.
More generally, Musk admittedly works with a "move fast and break things" mentality. His companies have repeatedly broken environmental laws, labor laws, and securities laws. His companies often fail to deliver the features they promise and routinely miss deadlines they've committed to. They routinely violate contracts that they have agreed to.
Not doing business with them is a completely normal and unremarkable decision.
I used cursor extensively both at work and for personal use. Its harness is pretty darn good but I basically never used OpenAI models in it.
The first party cursor models (initially composer and now grok-cursor) and price very competitively for their quality and have very generous usage limits included in the base plans, and then per token pricing afterwards.
Occasionally I’ve used some of the Anthropic models in it but found their cost massively outweighed the benefits for the type of usage we do.
I was using Sonnet 4.5 and 4.6 for a while but the newer models are a lot more expensive at the per-token pricing.
Cursor does currently allow users to BYO OpenAI token. Given Cursor won't have access to OpenAI models for prompt engineering/evals though, there's definitely a question of whether they'd allow picking any new OpenAI models, assuming the BYOK option continues to exist at all.
Email to customers now "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.
All it shows is that openai is likely still subsidizing inference, sees no value in integration buried within other branded tooling competing with theirs.
I used mostly OpenAI for the last like 6 months in cursor. I got tired of how slow Claude models are and switched. I’m surprised usage is so low. Wonder why.
Spend a lot of time thinking about why individuals shouldn't trust large AI cos, now it's interesting getting to think about how the AI companies themselves don't trust each other. Capitalism is viciously cold and uncaring.
Not sure the larger point you're trying to make (maybe we're on the same page?), but it's a fact that AI has come from heavy long-term investments by the academic and public sectors, yet private companies are accruing most of the gains
back when the USSR was still a going concern, there were zero American leftists claiming that it wasn't really communist. that only started happening after they failed
> We are making this choice because we cannot be confident that SpaceX will use our technology within our terms of service, based on our experience with Elon Musk's companies violating contracts.
> After Musk acquired Twitter, now part of SpaceX, the company broke (opens in a new window) the terms of our contract (alongside many others).
I have never seen one large company throw another large company under the bus without any legalese filters like this before. Wondering if OpenAI site got pwned. Or maybe we in some strange, alternative reality. I doubt even their LLMs would be this frank.
The way I read the article was that they will continue providing their current models, but won't let them use any newer ones announced even between now and the end date. But I might be wrong.
> we’ve decided to hold the contract cancellation to the latest date we can while not providing future models to Cursor.
Cursor is like one of the least economical ways to use Claude and OpenAI, virtually nobody does it. You use Cursor for their autocomplete/composer + grok. For using Claude and Codex, you get immeasurably more bang for your buck just using them directly.
All good points. However, it's not a safe assumption that everyone operates that way. People probably use the flagship models because that's what they know. Sure, some probably just use the default, in this case Grok.
Also, it's a signal more than anything. OpenAI pulls out of Cursor. Then Anthropic and Google does as well. Then you no longer have "Cursor" you have Grok.
Grok doesn't have the best reputation in general, regardless if you're tech savvy or not.
We are making this choice because we cannot be confident that SpaceX will use our technology within our terms of service, based on our experience with Elon Musk's companies violating contracts.
The pot calling the kettle black. Look who's talking. A company that trained it's models illegally on so many copyrighted materials without compensating anyone involved.
For a company named "Open" AI, they've been hardly open at all and the slew of resignations of top tier execs in the company is really the proof.
Make no mistake, this isn't a moral decision. This is just killing the competition using morality as an excuse.
Codex cli is open source rust project (Claude Code is not). Whisper is an open weight ASR/STT model. GPT-OSS is an open weight LLM that can run locally and is still very capable. CLIP is an open source model used for image encoding. Just some of the open contributions from OpenAI
i don't see the point of cursor when vsc + codex extension does the same thing. i don't use autocomplete anymore (who writes code anymore? lol) any and anyone using grok unironically is also probably a moron.
i only tell VC type people or others from whom i want something "oh yeah grok is cool" if they mention it, because they're always sucking off Elon
this is a pretty clean test of whether people were buying cursor or just using it for convenient access to whichever model was flavor of the week.
i work in qa so i'm naturally wary of adding another critical dependency to a development workflow, if the model you actually want can just disappear because two vendor change ownership or start argueing then the wrapper has to be really good on its own.
cursors harness and its own models may be good enough that users stay, and if they do cursor has a moat at least. if everyone follows the openai models out the door, it will just become a polished ui.
Anthropic already banned xAI for very similar ToS violations earlier this year: https://x.com/kyliebytes/status/2009686466746822731?s=46
So this is OpenAI following suit after Musk admitted to distilling their models.
This was bound to happen once Cursor decided to sell itself to a competing model provider.
It'll be interesting to see if Anthropic applies their ban to Cursor, or if that datacenter deal they signed with Musk changes anything.
Your link doesn't suggest a violation of TOS, but rather a policy shift disallowing their major competitors to use their model for any use.
That thread you linked suggested they were using it with cursor as part of their development process, not for distilling.
>"Hi team, I believe many of you have already discovered that anthropic models are not responding on cursor. According to cursor this is a new policy anthropic is enforcing for all its major competitors."
Cursor might have framed it that way at the time, but 4 months prior Anthropic cut OpenAIs access under the same clause in the ToS: https://www.wired.com/story/anthropic-revokes-openais-access...
And a month before the OpenAI block, they blocked Windsurf for the same reason: https://www.forbes.com/sites/johanmoreno/2025/06/05/anthropi...
So its pretty clear the TOS had this clause prior to Jan 2026.
> if that datacenter deal they signed with Musk changes anything
It's hard to imagine a 1B / month deal wouldn't.
>Musk admitted to distilling their models.
what's the moat for AI labs like OpenAI and Antrhopic as they seek trillion IPOs. a lot of people said Chinese models also distill US LLM models. if it is this easy to do. how do US AI labs justify asking for trillion?
Well, they will stop releasing their latest models, and just attempt to eat all software themselves.
Otherwise, it will be competitors distilling, and the government getting upset.
It's hard for me to see a different future.
And, this really sucks.
If this thesis is true, I wonder what the final YC batch number might be.
I believe if this was true, we'd already be seeing vibed stuff succeeding everywhere, and pricing out incumbents, but I guess I'm not seeing this? Like at all?
Even if they have Astra or whatever completely to themselves, how do they deal with the -product- side? Or sales, or account management?
They've been focused on juicing model's coding capabilities, but it's absolutely -not- "gen ai" enough to be doing the whole thing in agents, even 3 or 5 years from now, if only because there's so much context we _can't_ give to these models, without reverse centauring ourselves with cameras and mics and oodles of compute everywhere, and so far that hasn't exactly been playing out the way the frontier labs and the singularity folk hoped it would (meta glasses? humane? really?)
Distilling merely saves an expensive part of the process: hundreds of millions. So even if distilling wasn't a thing, what justifies trillions ?
Anthropic wants you to think that models that distill from them would be worthless otherwise. It’s not true, it’s just one part of the process. The whole narrative that Chinese and other models are only good because they distill is nonsense.
Markets and rationality are sometimes just acquaintances
Oracle makes billions selling SQL when you can just use a free version. This is no different. Enterprise has enterprise needs. It's really not that complicated or irrational.
i'll be surprised if anthropic doesnt follow. cursors biggest advantage has been giving users a polished wrapper around whichever frontier model they prefer
running on someone else's server, you're basically giving them your weights.
Anthropic seems to have no issues: https://x.com/NotTomBrown/status/2093541294027280657?s=20
Remind me to set up public nitter so I can post a nitter link to whatever this is, and then I can read it.
Not being able to read about Musk doing some shit because Musk have done some shit to Nitter is just peak comedy.
Just go ahead and do set it up. You’ll be my hero (I don’t have mental capacity to research which jurisdiction works for this right now, but obviously not in US or EU.)
LOL, obviously Space-cursor-xAI-musk-imperial-holdings-inc is providing them compute. They either _can't_ cut them off, or they know not to bite the hand that feeds them.
Yes Anthropic’s chief compute officer said nice things about one of Anthropic’s chief providers of compute. Recall that compute constraints are one of their existential risks.
Lol, Anthropic runs a significant portion of their compute on xAI hardware, wtf are you even talking about trying to suggest this nefarious shit?
Some of you guys hate Musk so much that you say blatantly nonsensical stuff. Like, how can you be on HN and not have some idea about the xAI-Anthropic compute contract, or the fact that xAI took down Grok capabity so they could sell compute to Anthropic? And if Anthropic hated Musk/xAI as much as you are saying they do, why would they be paying xAI billions for compute? Think before typing.
> how can you be on HN and not have some idea about the xAI-Anthropic compute contract
Some of us are still employed mate
It's ironic how AI companies are re-inventing their own form of privately enforced copyright, lobby the government to ban foreign competitors that don't respect it etc., all while spending the last 5 years fighting tooth and nail against the copyright of the training material they're using.
If you can take any book and turn it into a model, because it's "transformative enough", and "AI learns just like a person does", then surely a model distilling another model is transformative and fair use.
They tied themselves into knots fighting the letter of the law, and now, when they need protection in the spirit of the law - that each creator deserves protection for their work - now we devolve to the law of the jungle. Maybe we'll even see LLM book curses.
Are you willing to denounce each and every Chinese company for equal amounts of IP theft as well then?
Companies are well within their rights to choose who they have business relations with. I don't see how this is 'privately enforced copyright',.
I may be a rare Cursor user around here, but I am saddened by this news. The thing that keeps me using it is precisely the ability to switch between models: I usually do brainstorm/planning with Sol or Terra, then let Composer (free) deal with the implementation. For certain tasks I use Claude, which tends to be more expensive but sometimes saves me time.
Using both OpenAI and Anthropic models, Cursor is cheaper than having two separate subscriptions, and their free tier of Composer is good enough. I will now have to reconsider my options.
It's very easy to achieve that and more with: Crush (or Opencode) + OpenRouter right from the terminal. Takes a minute to setup.
If you can explain how to make compaction not suck in opencode or Kilo Code, I would love to know how. I am really trying to make use of other harnesses but that's a major sticking point.
Never compact. Not automatically, not manually. Instead, let it write ledgers and checklists or handover prompt if necessary, and start fresh sessions as often as possible.
currenly I use pi with openrouter and is very nice I combine with a chatgpt sub that allows me to use it with pi. and when my usage runs out I turn to openrouter. Then same when I need anthropic models
JetBrains is a good option - works with any model/agent. Also they have very good semantic search for the codebase, which is exposed via MCP to the agents.
I like Cursor. Honestly the biggest reason is the decent WYSIWYG markdown editor.
AmpCode is not popular here but I recommend it, give it I try and you will not go back, except the price may a bit expensive.
It's always been clear to me that Cursor's business model of reselling others' APIs had its days numbered. Not necessarily because the providers would pull the plug, but because you wouldn't be able to compete with subsidized plans.
Cursor is already kind of useless for third party models unless you're willing to spend thousands of dollars. It's only worth it if you're going to use mostly grok/composer.
Alternate take: open source models and model routers are the real bang for your buck. Subsidized plan or not.
Agreed. Wrapping token spend is great for pumping your revenue numbers (congrats to them on $60B!), but doesn’t seem sustainable over the long term.
Big reason we built https://boxes.dev around the model harnesses (Codex + Claude Code), so you can bring your own subscriptions.
Surely the days of being able to “bring your own subscription” to a third party platform are also numbered?
I’m not so sure… Claude backtracked on this when they tried it
They played pigeon on discless videogame consoles a time or two, too.
How long have you been in tech, outta curiosity?
Edit: great answer.
> Edit: great answer.
What do you mean by that? If you're being sassy about a downvote then I'll remind you that when you reply to a comment, that person can't downvote your reply. Also this is an account with barely any comments or points, I'm pretty sure they can't downvote at all.
On one hand, yes, it doesn't make sense to give out subsidized tokens forever.
On the other hand, subscriptions create lock-in in a way that API pricing doesn't.
I think a more likely end is that subscription value decreases over time because API pricing gets more reasonable, but subscriptions stay because they are a good way of getting money out of people consistently.
Right, subscriptions are not only cheaper because they "create lock-in". They also let you do things like forecast demand and plan your capacity for it, borrow against it sometimes if you need to, and keep somewhat less around because your cashflows are predictable. It is also valuable because it keeps customers around but that's by no means the only reason recurring revenue gets a higher multiple.
This is why "subsidized tokens" is possibly a misnomer. Money at lower variance is worth more than the same money at higher variance. Not "subsidy" so much as reducing risk and passing some of that to a consumer.
I just told my boss to cancel my cursor subscription. My flow is basically Claude/opencode + kyde/vscode for por changes review
Shit, for most things, ive got a dev agent that reads change requests from a GDoc and communicates through email with me. I can develop in my mobile
What about openrouter though?
I will slightly disagree on a small slice: Their harness/instruction makes Opus/Fable so much better to work with.
Honestly, _this_ was their moat more than Composer to me.
Couldn't you just download cursor, not even pay for any account and hook up an Anthropic api key and get that functionality?
I dont know how cursor works. Everytime ive tried to use, its super buggy and has memory leaks that make my very quite PC sound like a jet engine.
You can just pirate Adobe's software and their marketcap is 115.88B.
Cursor charges the “Cursor Token Rate” for BYOK https://cursor.com/docs/models-and-pricing#cursor-token-rate
If the Opus/Fable harness is superior and it's written using Opus/Fable then it will cease to be a moat. Otherwise the premise for coding using LLMs is essentially untrue.
> Otherwise the premise for coding using LLMs is essentially untrue.
Why? Written by does not mean designed by etc. There's a lot more to it.
I've never seen a product blow another product out of the water so hard as Claude Code did to Cursor.
Cursor has a surprising amount of inertia, though, at least where I work.
I've recently moved to a company that uses Cursor, and I'm actually quite fond of it.
I basically never use the editor but the fact that there's a review UI for all the agent work is incredibly helpful.
For me, that makes the models much more usable. I've also been using GPT models a lot, as they're cheaper and less vomit inducing than Claudes text, so this is definitely bad news for me.
Claude code is a decent harness but then you have to use Anthropic models, which are good but no longer the best bang for the buck.
In particular Haiku 4.5 is rubbish, Anthropic don’t have anything in the cheap/fast part of the market.
For people fortunate enough to still be on a subscription instead of per-token billing this is less relevant, but their time will come.
The cursor model is quite nice because it allows you to switch between cheap and expensive models for different tasks
I think the point is that CC was the first general use, terminal based harness. I think. Now you can use PI, or OpenCode, or whatever, and use any model you like.
Now, I’m no expert, because I was late to the game and have only ever used Pi. But I guess Cursor is some product that’s tied to your IDE? If so, then yeah. That’s just too restrictive. I still love my IDE, but I don’t want it to be my harness too.
> Claude code is a decent harness but then you have to use Anthropic models...
Claude Code works with models from other providers too. Anthropic supports this. You can configure some Claude Code environment variables to switch: eg changing ANTHROPIC_DEFAULT_HAIKU_MODEL to point to GLM Flash or Luna, setting ANTHROPIC_BASE_URL to point to api.z.ai, and making ANTHROPIC_AUTH_TOKEN the API key for your alternative provider instead.
Some instructions here:
https://docs.z.ai/scenario-example/develop-tools/claude
That said, I've not actually tried this myself, opting to build my own harness instead. And you don't know what information Claude Code might be sending back to Anthropic about how you use competing models and which models you use. I don't know for sure that they do this, but after hearing about how they used steganography in the date of harness system prompts to identify the user's location, I don't entirely trust Claude Code anymore.
I saw that happen culturally. But never saw that much different in performance. And I really dislike terminal development. And their desktop app is really buggy.
The tide is turning though. I was using the old non agent view in cursor every day for about a year when I built my startup with it. Then Claude Code blew it out of the water. But since two month or so I much prefer Codex and Cursor with the agent view.
However, what I see cooking at Cursor is way more promising than the others. Things like full system understanding, their own forge, multi-repo support. I'd put them as way more visionary in the Gartner magic Quadrant ;)
Enterprise customers are not getting the subsidized subscriptions so for them the point is moot.
Token reselling stops working when providers literally won't work with you
Why do people keep insisting that subscription plans are subsidized? They are not. API prices are outrageous, designed to milk enterprise users. They probably have an 80-90% profit margin on subscription plans.
Inference is cheap; Dax once said on a podcast that Opencode has a close to 90% profit margin on the openweight models it provides inference for. Those models are close in size and spec to models from large labs.
What a shame for Cursor. I found it a great tool - I actually liked Composer 2.5; but now being bought by SpaceX has for a while left a bad taste - and they keep pushing Grok and now GrokBot.
I took the 50% discount instead of cancelling - but with this I'll likely not renew.
So sad to watch them go from great to mediocre in such a short time.
While this isn't great for users it seems like a pretty standard circling of the wagons in preparation for the next phase in the battle for frontier AI dominance.
Ironically, as a Cursor and Claude subscriber, but not an OpenAI subscriber, this will push me back to Anthropic. I don't feel good about giving OpenAI money directly, but I've found switching to GPT 5.6 Sol now and again in Cursor to be useful. Now I'll probably just forget about OpenAI models.
I think they are betting that Astra will be impossible to ignore.
For three months, until the next SOTA model drops. And with each improved model, the number of people who actually need it declines.
Exactly. Astra will likely get its Mythos moment, and then another Chinese lab releases its equivalent three months later while Anthropic releases a newer, better model.
There's a reason why most businesses don't hire PhDs, they dont need it. PhDs are cheap in many countries, arguably more affordable than a full time mythos lvl agent, yet businesses in India still don't hire 30k a year PhDs to run their businesses. Its simply not necessary. AI labs have lost the plot from a business perspective. If chasing ASI is the goal, that also seems stupid.. Putting all of your customers out of business with ASI doesn't really lead to great outcomes either.
They're commodity fetishism maxxing.
I'd need to have a problem I couldn't solve with current models for my eyes to wander.
Yeah I haven’t had a problem yet which was too difficult for Fable to solve. I do have problems which Sol does not do well on. But if I could get Fable-level intelligence faster/cheaper and less restricted, that’s all I need.
Fable is great minus it's annoying writing style. It's nowhere near as bad as Opus 5 but it's very "AI trope-y" even with heavy steering. I can spot the writing immediately.
What if it’s cheaper too?
Yeah not sure why everyone thinks they need a PhD lvl assitant to build their slop SaaS when an open source model that is 100x cheaper, and 3x faster and arguably still "PhD" level assitant could also build their slop SaaS.
Why would you possibly allow companies to push you? You are the customer and you should control which models you use.
Because one is better or worse? What else do you think they meant?
OP said:
> as a Cursor and Claude subscriber, but not an OpenAI subscriber, this will push me back to Anthropic
Instead of deciding based on the merits of the model, they are being pushed by the actions of the companies. You should be able to use the best model, or whatever model you like, independently of what any company thinks or does.
It’s OpenAI (and Anthropic I’m sure) realising that they’ve somehow ended up stuck in the lowest margin part of the stack (providing inference), and trying to get out of it
I don't see how it makes sense to use models outside of Grok or Composer in cursor unless you like the tooling to such degree that you're willing to pay out the nose for access to OpenAI and Anthropic to use with that tooling. And Codex is regularly scoring towards the top among harnesses, so you might as well use the best harness available for the given model. I'm happy with Grok and Composer in Cursor. If Cursor wants to add more models, they should host more open models.
Is there a good place for harness comparison/scoring?
https://www.harness-bench.ai/leaderboard.html
This provides a method, but the data looks stale and perhaps a bit thin compared to say, Cursor, or even AntiGravity data.
never heard of nanobot. how reliable is this benchmark?
I can’t speak to the veracity of the benchmarks but it appears their methodology is sound. Nanobot has 47k stars, fwiw https://github.com/HKUDS/nanobot
It has been more of an OpenClaw or Hermes alternative than a coding agent like OpenCode or Pi, so it’s likely to do well given less context bloat.
IMO it's not. It's benchmarking GPT 5.4 and Opus 4.6. It's also missing Claude Code... one of the most popular harnesses (the most?)
No Pi, no Aider either.
The reason it makes sense is because different models are better at different things. Allowing yourself to be locked in to a particular model by these companies is a bad idea and will lead to worse outcomes for yourself and the market as a whole.
Because I pay $60 per month and not using the included API tokens on other models is wasteful.
If you have such a world changing technology, why sell an API when you can capture all the value yourself? That's what every provider is asking themselves.
Cursor will be fine after this, as they have their own first party models and can potentially use open weight models via API or their own hardware, which have gotten good enough to Opus and Fable quality now for many use cases.
Being a pure play model provider is completely untenable right now. Coding agents own your demand; Nvidia, data centers, and DRAM manufacturers own your supply. You're getting squeezed by open weight models. Your only bet to making money off this thing is by locking your model down to your program to have some semblance of a moat. The fact that businesses can change providers with just the change of an API key is unsustainable. You have neither the "hard to switch" lock-in of a traditional software company nor do you have the near-zero marginal cost of production they have. Each token of inference costs you money, not to mention the training costs. At least locking the models down to your harness lets you get some sort of lock-in.
Cursor founder just tweeted that he was sad to see this ban and that OpenAI model represented around 5% of total usage.
Burn.
> we are giving the maximum notice provided by our contract.
Anyone else surprised that the maximum notice period allowed is about 3 months?
I would have thought 3 months is minimum notice period, with provisions to stop faster if any provable violation occurs.
It makes no sense that a contract would even provide a maximum notice period. Why would parties put that in? What benefit would either derive from restricting themselves to not getting earlier notice.
I read it as: we acted as quickly as we could once it became possible to do so, i.e., the change in control was completed. not that their special agreement carved out some maximum notice term.
Everyone is playing hardball. No one wants to be IBM in an IBM-Microsoft dance.
But this is kind of like banning GPU exports to China. It will encourage them to develop their own technology. Yes, it will take them longer, but in the long-run China will be more successful with their own GPU tech.
The same could be true of SpaceX/xAI
LOL
You had me until:
>The same could be true of a company run by a drug addict.
Elon Musk is nowhere near as capable as millions of intelligent developers in China.
He doesn't have to be. He just has to hire millions of intelligent developers in China.
Bummed, but I guess for me, doesn't mean cursor subscription ending though. Grok is a pretty amazing model for the speed and price on Cursor.
Does Grok have a subscription that competes with Claude Max $100 plan (genuinely curious)
> As AI capabilities advance, we also have a new level of accountability to ensure our upcoming model, Astra, is being used in accordance with our terms.
I think Sam Altman has played a clever business move here, they can use this excuse immediately after the Hugging Face leaks and it's not going to be dismissed willy-nilly. Whereas long term they probably wanted to end their partnership with Cursor anyway. I bet Dario will be happy about this too, but maybe the compute deal prevents him from doing similar anytime soon.
Very interesting times we live in...
Is it really an excuse? As the release points out, musk admitted under oath to distilling OpenAI’s models to produce xai models on an ongoing basis. Whatever you think about goose for the gander wrt IP theft, I wouldn’t voluntarily do business with a competitor who was actively and brazenly flouting my terms, would you?
It's amusingly ironic to talk about "IP theft" in the context of AI models.
Shorthand. I agree it’s hypocritical and theft isn’t the right word, but we make do with the words we have.
OpenAI's models are distilled from your HN posts
You can still use it via API keys (answering my own question from earlier), https://x.com/thsottiaux/status/2093515916076343774?s=46
The last time I tried an API Key in Cursor they limited it to the very basic Ask mode. No Agent Mode. So this is very much useless if still the case
Any tool that tightly couples you to an inference provider is a sub-par tool and destined to be forgotten.
> As AI capabilities advance, we also have a new level of accountability to ensure our upcoming model, Astra, is being used in accordance with our terms.
This move seems to protect Astra against distillation. I wonder whether Astra will be released before or after 11/12.
Before, but they say that they’re also not going to provide astra to cursor subscribers
My guess is that OpenAI now has a full blown Cursor equivalent in development (an expansion of Codex?) and will announce it in September. Otherwise it stands to lose those who are used to working with Cursor.
Yeah, the "go above and beyond for our users" implies they have a "better" version of Cursor coming. Probably integrating their new chip/microphone based product.
OpenAI acquired Ona in June https://openai.com/index/openai-to-acquire-ona/
Does this also apply to API endpoints, let’s say if you want to provide API keys in Cursor to Azure or Bedrock deployment of a GPT model?
I can't imagine it would.
Why maximum notice? Why not minimum notice? Elon doesn't deserve more than that.
Allowing the company or harness to dictate which models you use is crazy.
Honestly, can't say I miss my Cursor subscription at all, and I'm surprised they're still a going concern. Why would I want to use a proprietary VSCode fork with an identity crisis when I can use literally any editor with any agent and have about the same experience?
The integrated review UI is super helpful, but apparently nobody reviews code anymore except me.
> We know that the people most affected by this decision are the developers who rely on OpenAI models in Cursor. We care about their experience in this transition and we’re ready to go above and beyond to support them.
Do I smell a cursor-analogue from openAI?
Damn; Cursor has basically no value if Anthropic pulls the same move. Twitter Part 2 indeed.
Anthropic rents a lot of capacity from xAI, so I don't think they're in a position to do the same.
I’d be canceling for sure Grok is a huge joke and consistently destroys my Clojure files.
How so? A dozen very capable models out there. And the gap between frontier models from the two main US labs kept narrowing, at this point non existent.
The quality/price ratio is clearly superior when running Qwen, GLM, or Deepseek.
To make a product more exclusive won't hide the fact that product is becoming more and more irrelevant.
It was honestly just a matter of time, given all of the bad blood there is between Sam and Elon and Elon basically admitting under oath they were distilling other companies models.
I wonder if Anthropic will follow suit given their campaign against distillation or if they'll stay quiet given their reliance on SpaceX (among others) for hardware.
damn... cursor doesn't get as much love as claude code or codex (on here anyways), but honestly, it's gotten amazing in the last 6 months.
i loved switching between sol / fable / grok 4.5.
... and something tells me fable might not be available on cursor for much longer :/
Why would they give the maximum time instead of the minimum given violation has already occurred?
Because it doesn't matter anymore for current models and they want to show goodwill, probably.
This low key confirms Astra for November
They were going to do this regardless of SpaceX.
Not surprised, but honestly I don't feel like this is a huge loss for cursor. Models are so commoditized, and 99% of users don't need particularly fancy models. Between existing anthropic models, composer-2, honestly even grok models, and also the fact that many people just use 'auto' mode, I feel like many won't feel the impact for OpenAI's withdrawal.
Maybe if Claude joined along it would be more impactful. For better or worse though, Xai seems like its lining themselves up to handle being on their own for cursor.
At my dayjob we are big users of Cursor's Cloud Agents. We use them to trigger automated investigations customer support issues, alarms, feature requests etc. The Cursor editor is fairly meh in comparison, but this has become a big part of our workflow. And having anyone on our team manage them. What competing products are there we should try?
It's going to be tough for Origin, their code hosting platform, to follow in the footsteps of GitHub, if they can't be treated as a neutral party.
Elon Musk openly admitted to distilling OpenAI's models [1], which I think explains the "based on our experience with Elon Musk's companies violating contracts".
Last year, Anthropic cut off Windsurf the same way [2]. Fortunately, unlike Anthropic, OpenAI allows you to use your subscription with other harnesses, including Cursor.
Of course today, Anthropic is taking the "high road" and is OK with with SpaceX [3] including their models in Cursor, because they need SpaceX's compute.
[1] https://www.forbes.com/sites/antoniopequenoiv/2026/04/30/elo...
[2] https://techcrunch.com/2025/06/03/windsurf-says-anthropic-is...
[3] https://x.com/NotTomBrown/status/2093541294027280657
More generally, Musk admittedly works with a "move fast and break things" mentality. His companies have repeatedly broken environmental laws, labor laws, and securities laws. His companies often fail to deliver the features they promise and routinely miss deadlines they've committed to. They routinely violate contracts that they have agreed to.
Not doing business with them is a completely normal and unremarkable decision.
Wow, I don’t use cursor but that seems pretty bad for them.
Is there a way to work around this (like using your OpenAI limits directly?)
I used cursor extensively both at work and for personal use. Its harness is pretty darn good but I basically never used OpenAI models in it.
The first party cursor models (initially composer and now grok-cursor) and price very competitively for their quality and have very generous usage limits included in the base plans, and then per token pricing afterwards.
Occasionally I’ve used some of the Anthropic models in it but found their cost massively outweighed the benefits for the type of usage we do.
I was using Sonnet 4.5 and 4.6 for a while but the newer models are a lot more expensive at the per-token pricing.
Cursor does currently allow users to BYO OpenAI token. Given Cursor won't have access to OpenAI models for prompt engineering/evals though, there's definitely a question of whether they'd allow picking any new OpenAI models, assuming the BYOK option continues to exist at all.
Email to customers now "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.
It's basically not going to affect them at all. I don't know a single person that uses OpenAI models via Cursor. Saw another comment say it was 5%.
All it shows is that openai is likely still subsidizing inference, sees no value in integration buried within other branded tooling competing with theirs.
I used mostly OpenAI for the last like 6 months in cursor. I got tired of how slow Claude models are and switched. I’m surprised usage is so low. Wonder why.
Spend a lot of time thinking about why individuals shouldn't trust large AI cos, now it's interesting getting to think about how the AI companies themselves don't trust each other. Capitalism is viciously cold and uncaring.
At least capitalism isn't at the center of the most impactful technological revolution in human history.
Yeah, that’s mostly been driven by socialism
Not sure the larger point you're trying to make (maybe we're on the same page?), but it's a fact that AI has come from heavy long-term investments by the academic and public sectors, yet private companies are accruing most of the gains
The iPhone too, etc. These ppl, who think Capitalism as a religion in a couple of years will be singing another tune with the same reassurances.
The movie Goya’s Ghost explains this in great detail.
show me a communist country
back when the USSR was still a going concern, there were zero American leftists claiming that it wasn't really communist. that only started happening after they failed
did they even have the internet back then? how much did people know about other countries?
> We are making this choice because we cannot be confident that SpaceX will use our technology within our terms of service, based on our experience with Elon Musk's companies violating contracts.
> After Musk acquired Twitter, now part of SpaceX, the company broke (opens in a new window) the terms of our contract (alongside many others).
I have never seen one large company throw another large company under the bus without any legalese filters like this before. Wondering if OpenAI site got pwned. Or maybe we in some strange, alternative reality. I doubt even their LLMs would be this frank.
Sam and Elon have a long standing feud. This is more than just bad blood between companies.
Still, where's the legalese? I'd expect something like this on an individual's blog, not a company's.
They can skip the legalese because Musk had testified under oath that Grok was distilling OpenAI models while Elon was on the board of OpenAI.
https://www.theverge.com/ai-artificial-intelligence/921546/e...
And to announce it on a Friday night. Brutal.
Elon Musk openly admitted to distilling OpenAI's models during his lawsuit against OpenAI. [1]
[1] https://www.forbes.com/sites/antoniopequenoiv/2026/04/30/elo...
they’ll see an uptick in codex user count trying the new models and can build a narrative around that too
So Astra release Friday 13th November!?
The way I read the article was that they will continue providing their current models, but won't let them use any newer ones announced even between now and the end date. But I might be wrong.
> we’ve decided to hold the contract cancellation to the latest date we can while not providing future models to Cursor.
Models are now having much tighter integration with harness.
It's gunna to happen anyway.
We are sailing into the age of vertically integrated intelligence.
This is personal, not business.
I guess I’m on the right track.
https://news.ycombinator.com/item?id=49428835
If Claude pulls away from Cursor as well then Cursor is basically done.
Grok? Google?
I'm not a fan of Elon, nor do I use Cursor, but OpenAI and Anthropic aren't the only models.
Cursor is like one of the least economical ways to use Claude and OpenAI, virtually nobody does it. You use Cursor for their autocomplete/composer + grok. For using Claude and Codex, you get immeasurably more bang for your buck just using them directly.
I've been using Cursor with Grok 4.6 a lot over the last week or so. Grok Build as well. Both options are performing very well for me.
Ya I was slow to try Grok but I’m now happy with it.
All good points. However, it's not a safe assumption that everyone operates that way. People probably use the flagship models because that's what they know. Sure, some probably just use the default, in this case Grok.
Also, it's a signal more than anything. OpenAI pulls out of Cursor. Then Anthropic and Google does as well. Then you no longer have "Cursor" you have Grok.
Grok doesn't have the best reputation in general, regardless if you're tech savvy or not.
The pot calling the kettle black. Look who's talking. A company that trained it's models illegally on so many copyrighted materials without compensating anyone involved.
For a company named "Open" AI, they've been hardly open at all and the slew of resignations of top tier execs in the company is really the proof.
Make no mistake, this isn't a moral decision. This is just killing the competition using morality as an excuse.
Codex cli is open source rust project (Claude Code is not). Whisper is an open weight ASR/STT model. GPT-OSS is an open weight LLM that can run locally and is still very capable. CLIP is an open source model used for image encoding. Just some of the open contributions from OpenAI
Correct. It's killing the competition. Which is what you should do if you want to run a successful business.
helloooo
What happened to researchers embracing the open research culture of improving on the results of others?
Days until cursor announces open models… 3, 2, 1…
Seriously though, when the api providers start blacklisting cursor and they’re forced to swap to open weight models…
If that happens, what even is the difference between openrouter and cursor?
They already have open models.
https://cursor.com/docs/models-and-pricing
Well, one has a pretty popular harness.
sure?
You assume OpenAI doesn’t turn off access to OpenRouter too.
Strategically, OpenAI long-term will want to own the direct relationship with all their customers. As will most major AI labs.
Makes sense.
i don't see the point of cursor when vsc + codex extension does the same thing. i don't use autocomplete anymore (who writes code anymore? lol) any and anyone using grok unironically is also probably a moron.
i only tell VC type people or others from whom i want something "oh yeah grok is cool" if they mention it, because they're always sucking off Elon
pot calling kettle black
Feels like scam altam writes the openai blog posts himself. The way Musk is brought up feels very 'mean girl'y
OpenAI / Sam Altman is probably even shadier than musk
“Satan is probably even shadier than Beelzebub”
this is a pretty clean test of whether people were buying cursor or just using it for convenient access to whichever model was flavor of the week.
i work in qa so i'm naturally wary of adding another critical dependency to a development workflow, if the model you actually want can just disappear because two vendor change ownership or start argueing then the wrapper has to be really good on its own.
cursors harness and its own models may be good enough that users stay, and if they do cursor has a moat at least. if everyone follows the openai models out the door, it will just become a polished ui.
I use Cursor because I listened to YouTube interviews with the teams behind most of these products and I trusted their team the most.