How do I point Cursor at FreeModel? Three fields in a settings file

Setup

In Cursor, open settings and add an OpenAI-compatible provider:

{
  "openai": {
    "baseUrl": "https://freemodel.online/v1",
    "apiKey": "sk-your-key"
  }
}

The base ends at /v1. Cursor appends /chat/completions itself — adding it here gives you /v1/chat/completions/chat/completions and a 404.

Choosing a model inside Cursor

Cursor's model picker expects names it recognises. With a custom provider you supply one yourself, so put a tier alias in and let the router worry about it:

If Cursor insists on a recognised name, use a concrete id from /v1/models instead — the endpoint lists every id this gateway will accept.

If it does not work

Next

Set up in three steps

1. Step 1: Open Cursor's settings and find the models panel. Use the panel if the override should apply to this machine only; edit the settings file if you want a configuration you can keep and move between machines.
2. Step 2: Turn on the OpenAI override and paste the base URL https://freemodel.online/v1 with your sk- key. Leave the base at /v1, since Cursor appends /chat/completions and a base that already carries a path segment gives you a doubled one.
3. Step 3: Add the model name in the same panel, then select it before you start working. A saved provider is not an active one: if the answers look like they came from Cursor's own default model, the custom entry was never picked.

Getting a key takes one step and no payment: sign in to the console, open My → API Keys, and create one. It starts with sk-.

What a custom provider actually changes

The override applies to the requests Cursor sends to the OpenAI-compatible provider you configured, which means the model you added to the picker is the model that answers those. What the picker cannot tell you is depth behind a name: the 199 listed models put chat, image, video, speech and embeddings in one list[1], so the same key reaches far past the editor's own work.

A tier alias in that field is worth considering for the same reason it is worth considering anywhere else. fm-v1-standard is not a model name, it is a standing instruction to the router, so a slow provider loses the next request instead of costing you a timeout. The trade is repeatability: two identical prompts may not reach the same model.

If Cursor refuses a name before sending anything, that is local validation rather than a gateway error. Paste a concrete id from /v1/models and the same request goes through.

Cursor FAQ

Sources

Every third-party figure quoted on this site was read from the vendor documentation linked below, on the date stated. None of these links are affiliate links and none of these vendors paid to be listed. If a number here disagrees with the page it came from, the page is right — tell us and it gets corrected.

• OpenRouter — credit-purchase fees and free-model rate limits
• OpenRouter — rate-limit reference
• OpenRouter — public model list
• LiteLLM — licence, deployment options and supported models
• Cloudflare AI Gateway — features and pricing
• Vercel AI Gateway — features
• Portkey — plans
• Helicone — plans
• Model Context Protocol — specification
• OpenAI Python SDK — base_url and OPENAI_BASE_URL
• Anthropic Messages API — request format

Where the percentages come from

Every share quoted on this site is counted from the live model list, not estimated. Divide the two counts in the last column, round to the nearest whole percent, and you get the number in the first.

[1] Counted on 29 September 2026 from GET https://freemodel.online/v1/models, which needs no key and no account. Every row uses all 199 listed entries as the denominator, because the only thing the endpoint states for every entry is its modality. It no longer publishes per-model capability flags such as tool calling or reasoning, so this site quotes no shares for them. Fetch the list and recount whenever you want to check the arithmetic.