Using Cursor with your own API key: what it costs, October 2026
What one Cursor agent request costs on your own key, worked at each developer's list price and ours, and the four settings that point Cursor at it.
The short answer
With your own API key, Cursor bills you per token at the price of the model you pick, and a request costs its input plus its output tokens. At 30,000 tokens in and 1,000 out, one agent request on Claude Sonnet 5.5 is $0.039 here and $0.07 at Anthropic's list price.
Prices
Cursor's API key setting speaks the OpenAI format, so any model that answers in it works, Claude included. These are five of them.
| Model | Input | Output | List input | List output |
|---|---|---|---|---|
| Claude Sonnet 5.5 | $1.20 | $3.00 | $2.00 | $10.00 |
| GPT-6.1 Sol | $0.60 | $3.00 | $2.00 | $10.00 |
| GPT-5.5 | $1.50 | $9.00 | $5.00 | $30.00 |
| DeepSeek V4 Pro 0813 | $0.75 | $1.188 | $1.32 | $3.96 |
| Gemini 3.8 Flash | $0.45 | $2.25 | $0.75 | $3.75 |
USD per million tokens. Ours from the live catalog; list is the price on the developer's pricing page, the read date on each model page.
One request, worked
An agent request in Cursor carries the conversation, the files it read and the tool results. Take 30,000 input tokens and 1,000 output tokens. That is an assumption, not a measurement; your requests show their real counts on the console's Usage page.
| Model | At list price | Here | Saving |
|---|---|---|---|
| Claude Sonnet 5.5 | $0.07 | $0.039 | 44% |
| GPT-6.1 Sol | $0.07 | $0.021 | 70% |
| GPT-5.5 | $0.18 | $0.054 | 70% |
| DeepSeek V4 Pro 0813 | $0.0436 | $0.0237 | 45% |
| Gemini 3.8 Flash | $0.0263 | $0.0158 | 40% |
30,000 input + 1,000 output tokens, each at its rate per million. No tokens are counted at the cached rate, so this is the upper end.
When a response reports cached input tokens, they bill at the model's cached rate instead, which is a tenth of the input rate or less on these models. A request that repeats a long prefix costs less than the table.
A day of use
Three sizes of day, each made of the request above. The sizes are assumptions; change the count and the table scales with it.
| Day | Claude Sonnet 5.5, list | Claude Sonnet 5.5, here | GPT-6.1 Sol, list | GPT-6.1 Sol, here | GPT-5.5, list | GPT-5.5, here | DeepSeek V4 Pro 0813, list | DeepSeek V4 Pro 0813, here | Gemini 3.8 Flash, list | Gemini 3.8 Flash, here |
|---|---|---|---|---|---|---|---|---|---|---|
| Light, 50 requests | $3.50 | $1.95 | $3.50 | $1.05 | $9.00 | $2.70 | $2.178 | $1.1844 | $1.3125 | $0.7875 |
| Steady, 150 requests | $10.50 | $5.85 | $10.50 | $3.15 | $27.00 | $8.10 | $6.534 | $3.5532 | $3.9375 | $2.3625 |
| Heavy, 300 requests | $21.00 | $11.70 | $21.00 | $6.30 | $54.00 | $16.20 | $13.068 | $7.1064 | $7.875 | $4.725 |
Each request is 30,000 input + 1,000 output tokens, none cached.
Setup
- Cursor Settings (Cmd/Ctrl + ,) → Models → API Keys.
- Turn on OpenAI API Key and paste your UplinkSwitch key from the console.
- Turn on Override OpenAI Base URL, enter
https://api.uplinkswitch.com/v1and click Verify. - Under Models, add the ids you want and pick one in chat.
claude-sonnet-5-5
gpt-6.1-sol
gpt-5.5
deepseek-v4-pro-0813The full steps are in the Cursor guide.
What the key covers
- The override covers every model in the picker. Switch it off to use Cursor's own.
- Tab completions always use Cursor's built-in models.
- TLS errors: Settings → Network → HTTP Compatibility Mode → HTTP/1.1.
Questions
What does a month cost? Multiply a row of the day table by your working days. Twenty steady days on Claude Sonnet 5.5 is $117.00 here and $210.00 at list, before any cached tokens.
Which model is cheapest? Per request, the one with the lowest input price, because input tokens are most of an agent request. Every chat model and its price is on the Models page.
Can I check my own numbers first? Re-price your last OpenAI or Anthropic bill on the savings page. New accounts start with $1 of free credit.