---
title: "Migrating from Legacy Models to Prediction Platform"
description: "Take advantage of our new Prediction Platform that provides real-time prediction on even more powerful AI models."
---

**[Understanding Multi-Model API](/pt-BR/use-cases)**

Leverage a single API to access multiple models with a scaling-ready architecture. Chain downstream models to integrate Avra's full intelligence into your operation.

Use Cases

**[Model Lifecycle Versioning](/pt-BR/integrate/model-lifecycle)**

Customers always retain control over when and how model upgrades impact their process—ensuring reliability while continuously benefiting from innovation and high-availability.

Explore

1. **Identify the available models for your Organization**

    Use the new **Get Organization Models** endpoint (see API Reference → Model Registry) to retrieve the list of models that are available for your Organization. This serves as a discovery mechanism, enabling applications, services or developers to understand which models they have access to, along with their identifying metadata.

    Example of models list:

    ```json
    [
      {
        "id": "08693cab-8830-4f66-83c8-90509b656794",
        "name": "Credit Score",
        "description": "A model for credit intelligence operations.",
        "type": "credit-intelligence"
      },
      {
        "id": "c98b4b8b-b0f0-428e-8d57-5e8de3456548",
        "name": "Embedding Model",
        "description": "Produces vector embeddings for business intelligence vectors that capture everything important about a Brazilian company.",
        "type": "embedding"
      }
    ]
    ```

2. **Pick a version from the chosen model**

    Use the new **Get Model Versions** endpoint (see API Reference → Model Registry) to retrieve all registered versions of a given model, along with the model's metadata. This enables you to present a version picker in UIs, validate a target version before running jobs or automatically select the active version for production.
    Model versions can be either **active** or **inactive**. A version may be inactive because a newer version has been promoted, or because it is still in the process of deployment. Only active versions are available for predictions. If the version you want to use is not active, please contact your Avra account manager.

    ```json
    {
      "model": {
        "id": "08693cab-8830-4f66-83c8-90509b656794",
        "name": "Credit Score",
        "description": "A model for credit intelligence operations.",
        "type": "credit-intelligence"
      },
      "versions": [
        {
          "model_version_id": "cf9b3583-0c30-40f4-a1eb-85c3c266945b",
          "is_active": false,
          "version": 1
        },
        {
          "model_version_id": "257ac7de-e74a-40fc-8a21-a18fdaa94d6f",
          "is_active": true,
          "version": 2
        }
      ]
    }
    ```

3. **Start Using the Prediction Platform**

    Use the new **Perform Prediction** endpoint (see API Reference → Model Registry) to perform downstream tasks on a specific model's version and receive results in real-time.

    > **Info**
    >
    > Remember: each model exposes its own input and output schemas tailored for your use case. You can always test the endpoint during implementation to validate the integration.