Skip to main content
Use DeepSeek models in the prompt playground and evals.
By adding this integration, your data may be sent to DeepSeek for certain actions within Arize AX (e.g., prompt playground) and your account may be billed for usage.
DeepSeek stores and processes data in the People’s Republic of China. Review DeepSeek’s privacy policy before you send data.
DeepSeek can’t be used as the model that powers Alyx. Alyx can still create tasks and experiments that use your DeepSeek integration.

​
Add the Integration Using the Arize Skills

Add a DeepSeek integration from your coding agent using the Arize Skills:

Create a DeepSeek integration named 'my-deepseek' with my API key sk-...

​
Add the Integration from Arize AX

​
Select DeepSeek from the Providers List

Navigate to Settings -> AI Providers and select DeepSeek.
The AI Provider integrations tab with the DeepSeek tile in the Get Started grid

The AI Provider integrations tab

​
Fill Out Your Integration Details

The DeepSeek integration details pane with the data location notice, name, API key, custom models, and test integration fields

The DeepSeek integration configuration pane

Enter a name and your DeepSeek API key.

​
Advanced Settings

The AI integration details tab

The Advanced Settings configuration pane

Expand the Advanced Settings section to configure the following:
  • Supports Function Calling - Turn this on to allow the integration to use function calling features. This is on by default.
  • Authorized Org(s) - Configure which organizations have access to this integration.
  • Authorized Space(s) - Configure which spaces have access to this integration.
At least one organization and space combination is required. The form selects all available organizations and spaces by default. Only Account Admins can use account-wide scope; other administrators grant access to specific organizations and spaces.

​
Add the Integration

Finally check the box to agree to the terms and conditions, then select the Add Integration button to create the integration.

​
Choose a DeepSeek Model

The model picker lists the DeepSeek models your API key can use, such as deepseek-v4-pro. DeepSeek models think before they respond by default. To change this in the prompt playground, set Reasoning Effort to Off, low, high, or max.