# Use BigModel.io with Cursor

> Use GLM, Kimi, DeepSeek and other BigModel.io models in Cursor chat and agent through the OpenAI-compatible base URL https://bigmodel.io/v1, with limits and fixes.

Web version: https://bigmodel.io/docs/cursor · Updated 2026-10-18

Cursor lets you add your own provider keys for chat models. With an OpenAI key slot, a base URL override and a custom model name, its chat and agent can call BigModel.io models over `https://bigmodel.io/v1`.

| Field in Cursor | Value |
|---|---|
| OpenAI API Key | your BigModel.io API key (`sk-...`) |
| Override OpenAI Base URL | `https://bigmodel.io/v1` |
| Custom model name | a model ID from the [model catalog](https://bigmodel.io/models), e.g. `glm-5.3` |

> Cursor's official API keys page documents custom provider keys but no longer mentions the **Override OpenAI Base URL** toggle. The steps below follow community reports from current Cursor releases. If your version has no override field, Cursor cannot be pointed at BigModel.io.

## What uses your key

- **Chat and agent** requests with the models you add use your key and are billed to your BigModel.io balance.
- **Tab completion** and other Cursor features keep using Cursor's own models and plan. Custom keys apply to chat models only.
- **All requests go through Cursor's servers** before they reach BigModel.io, so the base URL must be publicly reachable. `https://bigmodel.io/v1` is; a `localhost` proxy is not.

## Before you start

1. Create a key in [API Keys](https://bigmodel.io/keys) in a group that serves the models you want. Cursor sends Chat Completions requests, so every group that serves text models works.
2. Check the key with a minimal request:

```bash
curl https://bigmodel.io/v1/chat/completions \
  -H "Authorization: Bearer $BIGMODEL_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"model":"glm-5.3","messages":[{"role":"user","content":"ping"}]}'
```

## Configure Cursor

Menu names change between releases; look for the same fields if yours differ.

1. Open **Cursor Settings → Models**.
2. Under **API Keys**, paste your BigModel.io key into **OpenAI API Key**.
3. Turn on **Override OpenAI Base URL** (community-reported, directly under the OpenAI key) and enter `https://bigmodel.io/v1`. Do not add `/chat/completions`.
4. Choose **Add model** and enter the exact model ID, for example `glm-5.3`, `kimi-k3` or `deepseek-v4-pro`.
5. Enable the new model and select it in the chat or agent model picker.

While the override is on, requests for Cursor's built-in OpenAI model names also go to BigModel.io and may fail. Turn off the built-in models you do not need, or turn the override off when you want them.

## Choosing a model

Use IDs exactly as shown on the model page: lowercase with dots (`glm-5.3`, `kimi-k3`, `deepseek-v4-pro`, `minimax-m3`). Agent mode needs reliable tool calls; test with a short task before a large refactor.

## Verify the setup

Select your added model in the chat panel, send "ping", then open [Usage](https://bigmodel.io/usage). The request should appear with the model ID you entered. If it does not, Cursor answered with one of its own models.

## Troubleshooting

**Verification fails when you save the key.** Cursor may verify with one of its default OpenAI model names, which your key's group does not serve. Add your own model, keep only that one enabled, then verify again. A chat request that shows up in Usage is the real test.

**503 `api_error`: `No available accounts: no available accounts supporting model: ...`.** The model name in Cursor does not match a model your key's group serves. Copy the ID from the model page exactly.

**404 on every request.** The base URL is wrong. Use `https://bigmodel.io/v1`, not `https://bigmodel.io` or `.../v1/chat/completions`.

**401 `INVALID_API_KEY`.** The key was pasted with spaces, was deleted, or belongs to another account. Copy it again from [API Keys](https://bigmodel.io/keys).

**403 `INSUFFICIENT_BALANCE` or 429 `API_KEY_QUOTA_EXHAUSTED`.** Top up in [Billing](https://bigmodel.io/purchase) or raise the key's quota. A 429 `rate_limit_exceeded` names the key window (5h, 1d or 7d) that is used up.

**Built-in models stopped working.** The override sends all OpenAI-model requests to BigModel.io. Turn it off when you want Cursor's built-in models.

**Tab completion does not use BigModel.io.** Expected. Tab always runs on Cursor's models.

## Related

- [Configure Base URL in Codex, Claude Code, Cursor and Cline](https://bigmodel.io/blog/configure-base-url-codex-claude-code-cursor)
- [Cline](https://bigmodel.io/docs/cline) for a VS Code agent with a dedicated OpenAI Compatible provider.
- [Continue](https://bigmodel.io/docs/continue) for an open-source assistant whose chat runs fully on your own key.
- [Claude Code](https://bigmodel.io/docs/claude-code) for a terminal agent on the Anthropic protocol.