Cursor’s Models settings accept an OpenAI key and a base URL of your choosing. Flip two switches, type one id from the models page, and an uncensored model answers in the chat panel, in Ask and in Agent mode.
What you need
- Cursor, any recent build. Everything below happens in its Models settings.
- A key from the console. It begins with
sk-unr-. - One or more ids from the models page.
- Nothing to install if you only want to hear a model first: /chat runs the same catalog in the browser, three free turns a day, no account.
Find the settings
Open Cursor Settings with Cmd/Ctrl + , and go to Models. Below the list of models you will find the key field and the base URL field.
Three switches
- Enable OpenAI API Key and paste the key from the console into it.
- Enable Override OpenAI Base URL and type
https://api.unrestrictable.ai/v1. - Press Save. On older builds the same button reads Verify.
- In the Models list, add each id you plan to use. Every id on the models page is accepted:
glm-5.3-uncensoredglm-5.3-flash-uncensoreddeepseek-v4-pro-uncensorednemotron-3-ultra-550b-a55b-uncensoredThen choose one of them in the chat panel. Ask works, and so does Agent.
Each field, explained
- OpenAI API Key
- the whole sk-unr- keywhat Cursor attaches to every request as the credential
- Override OpenAI Base URL
- https://api.unrestrictable.ai/v1points those requests at this API rather than OpenAI; keep the trailing /v1
- Models
- the ids the picker offersone entry per id you want to use, spelled as the models page spells it
Which id to add
Cursor speaks the OpenAI format, and each build on the models page replies in it. Enter the id as printed, glm-5.3-uncensored for example, never the display name. In Agent mode a request carries the whole conversation plus every file Cursor opened and every tool result, so input tokens dominate the cost, and the build with the lowest input price costs least per turn.
| Model | One Agent turn |
|---|---|
| GLM-5.3 Uncensored | $0.01 |
| GLM-5.3 Flash Uncensored | $0.0047 |
| DeepSeek V4 Pro Uncensored | $0.04 |
| Nemotron 3 Ultra 550B Uncensored | $0.05 |
| GLM-5.3 Uncensored | Uncensored | $0.282 | $0.846 |
| GLM-5.3 Flash Uncensored | Uncensored | $0.1053 | $0.3685 |
| DeepSeek V4 Pro Uncensored | Uncensored | $0.97 | $1.94 |
| Nemotron 3 Ultra 550B Uncensored | Uncensored | $1.0913 | $6.402 |
Confirm from a terminal
First, list what the key can reach:
curl https://api.unrestrictable.ai/v1/models -H "Authorization: Bearer sk-unr-live-…"Next, send a single request shaped like Cursor's. With -i the headers print too: one carries the request id, and on a non-streamed reply another carries the charge in USD.
curl -i https://api.unrestrictable.ai/v1/chat/completions \ -H "Authorization: Bearer sk-unr-live-…" \ -H "Content-Type: application/json" \ -d '{"model": "glm-5.3-uncensored", "messages": [{"role": "user", "content": "Reply with the single word ok."}]}'Back in Cursor, pick one of your ids and send a message. It appears on the console’s usage page with its token counts and its charge.
How the bill works
- Each request is billed per token at that model’s price and drawn from your prepaid balance.
- If a reply reports cached input tokens, those are charged at the model’s cached rate rather than its input rate, so a long chat that keeps the same prefix gets cheaper per turn than a new one.
- Cursor’s API key docs say Teams and Enterprise plans count own-key requests toward Cursor’s usage and add Cursor’s own per-token fee, which Cursor bills, not us. Individual plans are not counted.
When it fails
401 invalid_api_key: the key is wrong or has been revoked. Paste it again from the console.404 model_not_found: the picker holds an id that is not on the models page. Compare the spelling.402 insufficient_balance: not enough balance for this request; top up in the console.402 monthly_limit_reachedis the monthly spend cap you set on the key.403 scope_denied: the key is scoped to some models, and this one is outside that list.429 rate_limited: the key’s rate limit was reached.retry-afteris in the response.- TLS errors: in Settings → Network, set HTTP Compatibility Mode to HTTP/1.1.
Questions
No. Tab runs on Cursor’s own models. The key and base URL cover the chat panel, Ask and Agent only.
Yes. Your ids join the same picker as Cursor’s own models, and you choose per conversation.
Yes. Agent sends tool calls in the OpenAI format, and the builds whose base model supports tools handle them; each model page says whether it does.
One strong build for the edits, one fast build for the questions between them. The table above prices one Agent turn on each at today’s catalog price. See the models page.
Every model above is on the uncensored model list with its live price per million tokens. Per-token pricing sets the whole catalog side by side, and the family directory groups the builds by base model.