# Use BigModel.io with Dify

> Add BigModel.io models to Dify with the OpenAI-API-compatible plugin, API endpoint URL https://bigmodel.io/v1, the right context size and Tool Call function calling for agents.

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

Dify adds third-party models through the **OpenAI-API-compatible** model provider plugin. Each BigModel.io model is added once, with its own context size and abilities. Requests go to `/v1/chat/completions`, which every text group serves.

| Field in Dify | Value |
|---|---|
| Model Type | LLM |
| Model Name | the exact model ID, e.g. `glm-5.3` |
| API Key | your BigModel.io API key (`sk-...`) |
| API endpoint URL | `https://bigmodel.io/v1` (must include `/v1`) |
| Completion mode | Chat |
| Model context size | the model's context window from its model page (the default 4096 is too small) |
| Upper bound for max tokens | the model's max output from its model page |
| Function calling | Tool Call, for models with tool calling (not "Not Supported") |
| Vision support | on only for models with image input |

Only workspace owners and admins can add model providers.

## Add a model

1. Open **Settings → Model Provider**.
2. Find **OpenAI-API-compatible** and install it if it is not installed yet.
3. Click **Add Model** on the provider.
4. Fill the fields from the table above. Enter the endpoint URL with `/v1` and without `/chat/completions`; the plugin appends `chat/completions` itself.
5. Set **Structured output** and **Stream function calling** only if the model supports them.
6. Click **Save**. Dify makes a test call before saving.

Repeat for each model you want, for example `kimi-k3` and `deepseek-v4-pro`. Then pick the model in your app's model settings.

## Verify the setup

Saving the model already runs a test call. To double-check, create a Chatbot app, select the model, send "ping" in the preview, and confirm the request appears in [Usage](https://bigmodel.io/usage).

## Troubleshooting

**Save fails with 404.** The endpoint URL is missing `/v1` or includes `/chat/completions`. Use `https://bigmodel.io/v1`.

**Save fails with 503 `api_error`: `No available accounts: no available accounts supporting model: ...`.** The Model Name is misspelled or not served by your key's group. Copy the ID from the model page.

**Save fails with 403 `INSUFFICIENT_BALANCE`.** Your balance is $0 or below. Top up in [Billing](https://bigmodel.io/purchase).

**Long conversations are cut short or error out.** Model context size is still 4096. Edit the model and set the real context window.

**Agents never call tools.** Function calling is set to "Not Supported". Edit the model and choose **Tool Call**, and use a model with tool calling.

**429 `rate_limit_error` during workflow runs.** Parallel nodes exceed your concurrency limit ("Too many pending requests" or "Concurrency limit exceeded for user"). Reduce parallelism or retry. A 429 `API_KEY_QUOTA_EXHAUSTED` means raise the key's quota in [API Keys](https://bigmodel.io/keys).

## Related

- [n8n](https://bigmodel.io/docs/n8n) for automation workflows.
- [LobeChat](https://bigmodel.io/docs/lobechat) for a chat UI.
- [Tool calling](https://bigmodel.io/docs/tool-calling) and [Structured output](https://bigmodel.io/docs/structured-output) for model abilities.