# Region-Specific Model Inference

The region-specific model inference feature allows you to route inference requests to specific geographic regions by using region-specific Model IDs when calling models. This meets requirements for data compliance, reduced access latency, or region-based billing.

## Feature Overview

The platform provides region-specific inference endpoints for certain models. Each region has independent Model IDs, vendor routing, and billing prices. You don't need to modify your existing calling logic—simply pass the corresponding region's Model ID in the `model` field to enable region-specific inference.

When calling a standard Model ID, the default inference region is global, and the platform will select the most appropriate region based on the access point. When calling a region-specific Model ID, the platform forwards the request to an inference node within that region.

The specific supported regions are displayed in the Model Marketplace. More model series and regions will be supported gradually. If you have special requirements, please contact customer support.

## Viewing Inference Regions in the Model Marketplace

### List Filtering

In the Model Marketplace list page, you can filter models that support specific regions using the "Inference Region" filter in the left sidebar.

### Viewing Model Details

Enter the model details page, and in the model capabilities section, you can see the "Inference Region" field, which displays the list of supported inference regions for that model.

If a model supports multiple regions, the pricing module at the bottom of the details page will display: Inference Region, Model ID, and Model Price.


## Calling via API with Region Specification

Region-specific inference does not introduce new API parameters. You only need to pass the region's corresponding Model ID in the `model` field.

Region-specific inference uses the same calling method and model protocol as standard models.

### Default Routing Call

When using a standard Model ID, the request is handled by default routing.

Example using Chat Completion protocol:

```json
{
  "model": "deepseek-v4-flash",
  "messages": [
    { "role": "user", 
      "content": "Hello" }
  ]
}
```

### Region-Specific Call

When using a region-specific Model ID, the request is routed to the corresponding region's access node.

Example using Chat Completion protocol:

```json
{
  "model": "deepseek-v4-flash-sg",
  "messages": [
    { "role": "user", 
      "content": "Hello" }
  ]
}
```

### Calling Notes

- Authentication, rate limiting, streaming responses, and other calling methods remain unchanged.
- Billing is executed according to the region-specific price corresponding to the actual Model ID passed.
- Region-specific Model IDs can be obtained from the "Inference Region" module on the Model Marketplace details page.


## Billing Notes

- When calling with a standard Model ID (default routing), billing is based on the standard model price.
- When calling with a region-specific Model ID, billing is based on the price configured for that region's model.
- Prices for the same model may vary across different regions. Please switch regions on the Model Marketplace details page to view corresponding prices.
- Region-specific models participate in the Consumption Center billing and model price table as independent models. The Model ID in the bill matches the Model ID you passed during the call.

## Viewing Region Information in Model Logs

Logs generated by region-specific model calls are recorded normally in the model logs. You can:

- Use the "Model ID" filter to filter call records for specific regions by region-specific Model ID.
- Region-specific models participate in log recording as independent models. Historical logs are not affected by the release or deprecation of region-specific models.