I’m trying to switch to a model connected via my own API key, but Cursor shows “Get more usage to switch models” and blocks the switch. This never happened before.
I have a BYOK / custom API key set up and need to use that API-backed model. The model picker / Ctrl+/ switch is blocked by this usage prompt. Previously I could switch freely to my API key models without hitting this gate.
Could you clarify whether this is expected under the new two-pool / usage-based billing (Other Models pool exhausted, on-demand not enabled), or whether BYOK models should still be selectable independently of Cursor’s included usage?
Hey, thanks for the report. Answering your question directly: on the individual Pro plan, BYOK/custom models via your API key shouldn’t be blocked by Cursor included usage. Requests sent with your own key bypass Cursor limits. Seeing the whole model picker replaced with Get more usage to switch models and not letting you pick a BYOK model isn’t expected behavior, and it’s not caused by anything in your settings.
A couple things that can help right now:
If a BYOK model was already selected before this screen showed up, it should keep working. Only the UI model switching is blocked, not the requests using your key.
To unlock the picker, go through Get More Usage and enable or raise your on-demand limit, or wait for the cycle reset. After that, model selection should be available again.
Small note: on token-priced Team/Enterprise plans, BYOK requests are billed at the Cursor Token Rate and count under Other Models. This doesn’t apply to individual Pro. Your BYOK traffic is fully exempt.
I’ve passed the locked picker issue to the team. I can’t share a timeline yet, but I’ll reply in the thread if there’s an update.
Hey, I can’t share an exact date yet. I’ll reply in this thread when there’s an update.
But you don’t need to wait for a real fix to use your custom model again. There’s a workaround you can use right now. Go to Get More Usage and enable or increase the on-demand limit. That will unlock the model picker, and you’ll be able to select your BYOK model again.
Important note: on individual Pro, your requests using your own API key are fully exempt from Cursor usage. So even if on-demand is enabled, there won’t be any charges for the custom model. On-demand is only needed to unblock the picker, not to pay for your key. Your API key traffic doesn’t count toward Cursor billing.
If you don’t want to touch on-demand, you can also wait for the cycle reset. After that, the picker should unlock too.
Also, if a BYOK model was selected before this screen showed up, requests through it should keep working. Only switching in the UI gets blocked, not usage with your key.
Let me know if the picker still doesn’t unlock after enabling on-demand.
Having the same issue, open AI API key is enabled, but it replaced whole models selector with “Get more usage” warning not allowing me to select any model at all.
Hey @Arthur_Bright, thanks for the report. This looks like the same issue described in that thread: when you run out of included usage, the whole model picker gets replaced with the Get more usage to switch models CTA, and you can’t select a BYOK model. This isn’t intended behavior and it’s not related to your setup. We’re tracking it, but I can’t share a timeline yet.
What you can do right now to unblock the picker:
Go to Get More Usage and enable or raise your on-demand limit. After that, model selection should be available again. Or wait for your cycle reset, the picker will also unlock then.
If a BYOK model was selected before this screen showed up, requests using your key should keep working. Only switching in the UI is blocked, not the traffic via your key.
Important: on individual Pro, requests using your own API key are fully exempt from Cursor usage, so enabling on-demand won’t charge you for the custom model. It’s only needed to remove the picker lock.
One quick question: what plan are you on, individual Pro or Team/Enterprise, and what Cursor version? On token-priced Team/Enterprise, BYOK requests are billed at the Cursor Token Rate and show under Other Models, so behavior can differ. Let me know if the picker unlocked after enabling on-demand.
@Arthur_Bright, your concern makes sense, and it’s totally fair. Let me clarify a couple things to reduce the risk of accidental charges.
On Individual Pro, requests made with your own API key are fully exempt from Cursor usage. Turning on on-demand by itself doesn’t charge anything. You only get charged if you actually send a request through a Cursor-billed model beyond your included usage. As long as you’re using your BYOK model, the cost stays at zero, even with on-demand enabled. On-demand is only needed to unblock the model picker, not to pay for your key.
To be extra safe against accidentally switching into on-demand when changing models, set a low spend limit in Dashboard → spend limit. Then even if you accidentally pick a Cursor model, spending will hit the cap you set. As an alternative, you can wait for your cycle reset, and the picker should also unlock.
On the main issue, the whole model selector being blocked by the Get more usage CTA and not letting you pick a BYOK model is not intended behavior and isn’t related to your setup. We’re tracking it. I can’t share a timeline yet, but I’ll reply in the thread when there’s an update.
Let me know if the picker unlocked after you enabled on-demand.
This latest updates make user experience way worse, and also permanent non removable “Usage limit reached” banner is pretty distracting and take useful space.
Maybe anyone has a custom extension workaround to hide this banner and restore the models menu, since official update probably would take a lot of time and looks like they just make UX worse?
Pro+ individual, both included pools at 100%. On-Demand is already enabled with a $5 limit and only $0.57 used, but the model picker is still completely locked behind “Get more usage to switch models”. I cannot switch models at all, including with Ctrl+/. Reloading, signing out/in and changing the on-demand limit did not fix it.
@Fernando_Gurkievicz, thanks for the details, this is exactly what we need. If on-demand is already enabled and no limit is selected, but the picker is still fully locked (and reload or sign out then sign in don’t help), that’s not expected and it’s not related to your settings. It’s the same issue being discussed in the thread, and we’re tracking it.
To understand why on-demand didn’t unlock the picker in your case, please send:
Request ID (top right of the chat > Copy Request ID)
your Cursor version (Cursor > About)
a screenshot of the picker with the Get more usage CTA and a screenshot of the Dashboard showing usage and on-demand
If a BYOK model was selected before this screen appeared, requests via your key should keep working. Only switching models in the UI is blocked, not the traffic via your key.
@Arthur_Bright, I wouldn’t recommend hiding the banner with a third-party extension. It won’t restore the model menu properly and it can break the UI after updates. The two correct workarounds are: enable or increase the on-demand limit (on Individual Pro, requests via your own API key are fully exempt from Cursor usage, so enabling on-demand by itself won’t create charges), or wait for the cycle reset. For the stuck picker even with on-demand enabled, I’m collecting the info above and I’ll post an update in the thread when we have one.
I enabled On-Demand Usage because Cursor would no longer let me switch models after reaching my usage limits. However, I am only using models through my own API keys.
After enabling On-Demand, Cursor has already charged $0.81 against my On-Demand budget, while I am also paying the model provider for the API usage.
If I supply and pay for my own API key, why is Cursor still consuming On-Demand credits?
Can someone explain exactly which part of the request is billed by Cursor when BYOK is enabled? It feels like I’m potentially paying both Cursor and the model provider for the same usage.
Hey @Rickard_Dahlqvist, let’s break it down, since the behavior depends on your plan.
On individual Pro: requests sent via your own API key are fully exempt from Cursor usage. Cursor won’t charge on-demand for them, the model cost goes only to your provider. If you’re on Pro and you see an on-demand charge, the request most likely didn’t go through the BYOK model and instead went through a Cursor model, for example the model picker switched to Auto. This is worth checking via the Request ID.
On token-priced Team/Enterprise: BYOK uses the Cursor Token Rate of 0,25 USD per 1M tokens. This is a Cursor infrastructure fee on top of the model, routing, search, processing, not the model itself. Inference cost is still billed by your provider, and Cursor only charges this small token rate, which shows up under Other Models / on-demand. So it’s not double-paying for the same thing, it’s two different parts. More details: Team Pricing | Cursor Docs and Models & Pricing | Cursor Docs
To figure out what happened with your 0,81 USD, can you share:
Which plan you’re on, individual Pro/Pro+ or Team/Enterprise
The Request ID for 1 to 2 charged requests right chat top right > Copy Request ID
Your Cursor version Cursor > About
With that, we can tell whether the request went through your key or switched to a Cursor model. If you want a review of a specific charge, it’s best to email [email protected], they can check your account details.