Scale inference service

Last update: September 07 2026Open in Markdown

To use inference service resources optimally, you can scale it depending on the load:

• configure a fixed number of inference instances — deployed model instances. For example, increase the number of inference instances when the request count grows or decrease it if the number of requests has dropped;

• or configure autoscaling. The number of inference instances will change automatically within the specified range depending on the number of requests and their processing time.

To scale an inference service, change the number of inference instances.

Change the number of inference instances

You can use a fixed number of inference instances or configure autoscaling.

Fixed number

1. In the Control panel, on the top menu, click Products and select Inference services.

2. Open the inference service page → Service tab.

3. In the Service autoscaling block, click Edit.

4. Open the Fixed tab and specify the number of inference instances.

5. Click Save. Changing scaling settings may take more than 10 minutes. During this time, the inference service will be unavailable.

Autoscaling

1. In the Control panel, on the top menu, click Products and select Inference services.

2. Open the inference service page → Service tab.

3. In the Service autoscaling block, click Edit.

4. Open the Autoscaling tab and set the minimum and maximum number of inference instances. The number of inference instances will change automatically within the specified range depending on the number of requests to the model and their processing time.

5. Click Save. Changing scaling settings may take more than 10 minutes. During this time, the inference service will be unavailable.