Gemini is a family of generative AI models developed by Google that is
designed for multimodal use cases. If you haven't used Gemini models on
Vertex AI before, see the [Generative AI
introduction](https://docs.cloud.google.com/vertex-ai/generative-ai/docs/learn/overview).

Key advantages of Gemini include the following:

- **Enhanced performance** : The latest large language models (LLMs), such as
  Gemini Flash 1.5, demonstrate better understanding across a range of
  natural language tasks than the AutoML text model. For
  more information, see the publicly available [technical
  report](https://arxiv.org/pdf/2403.05530) from the Gemini team.

- **Flexibility**: Gemini allows for both prompting (quick adaptation)
  and fine-tuning (deeper customization), catering to different project needs.
  This flexibility allows for rapid prototyping, testing, and deployment using
  prompting, with the option to fine-tune the Gemini model weights for
  optimal performance on specific tasks. Vertex AI offers both
  console-based fine-tuning and SDK and API options for programmatic control.

- **Multipurpose and multimodal capabilities**: Gemini offers the
  ability to process text, images, and other modalities. This approach enables consistent use of a single format and model across various tasks. This flexibility allows the process to be easily adapted for different applications, streamlining and accelerating development efforts.

Gemini supports most features available in AutoML text.
However, there are
[differences](https://docs.cloud.google.com/vertex-ai/docs/start/automl-gemini-comparison.md.txt), and the client
libraries don't support client integration backward compatibility. In other
words, you must plan to migrate your resources to benefit from Gemini
features.

If you are planning a new project, you should build your code, job, dataset, or
model with Gemini. This lets you take advantage of the new features and
service improvements as they become available.

## Recommended steps for migrating to Gemini

> [!NOTE]
> **Note:** AutoML Text is no longer supported.

Use the following recommended steps to update your existing code,
jobs, datasets, and models from AutoML text to Gemini.

1. Read about the major differences between Gemini
   and AutoML text at
   [Gemini for AutoML text
   users](https://docs.cloud.google.com/vertex-ai/docs/start/automl-gemini-comparison).

2. Review any potential changes in pricing (see
   [Gemini migration pricing](https://docs.cloud.google.com/vertex-ai/docs/start/migrating-to-gemini.md.txt#compare-legacy)).

3. Take inventory of your Google Cloud projects, code, jobs,
   datasets, models, and users with access to AutoML text. Use
   this information to determine which resources to migrate
   and ensure that the correct users have access to the migrated
   resources.

4. Review any [changes to IAM roles](https://docs.cloud.google.com/vertex-ai/docs/start/migrating-to-gemini.md.txt#changes-to-iam-roles), and then
   update service accounts and authentication for your resources.

5. Migrate your resources using either of these two methods:

   - [Prompting](https://docs.cloud.google.com/vertex-ai/generative-ai/docs/models/gemini-tuning#gemini-prompts)

   - [Fine-tuning](https://docs.cloud.google.com/vertex-ai/generative-ai/docs/models/gemini-tuning#fine-tuning-gemini)

6. View the [locations](https://docs.cloud.google.com/vertex-ai/generative-ai/docs/learn/locations)
   available for Gemini.

7. [Identify usage of
   AutoML text APIs](https://docs.cloud.google.com/vertex-ai/docs/start/migrating-to-gemini#identifying-api-usage) to help determine
   which of your applications use them
   and to identify the method calls that you want to migrate.

8. Update your applications and workflows to use Gemini.

9. Plan your request quota monitoring. See [Quotas and
   limits](https://docs.cloud.google.com/vertex-ai/generative-ai/docs/quotas).

## Gemini migration pricing

Migration is free. After migration, legacy resources are still available to use
in AutoML text until the service shuts down in June 2025. To avoid
unnecessary costs, shut down or delete legacy resources after you have verified
that your objects have migrated successfully.

## Gemini pricing compared to AutoML text pricing

Gemini pricing is generally cheaper compared to equivalent tasks in
AutoML text. Gemini pricing is determined by whether you're
using the model for prompt engineering only, fine-tuning only, or a combination
of both. For more information, you can compare [AutoML text
pricing](https://cloud.google.com/vertex-ai/pricing#automl_models) with [Gemini
pricing](https://docs.cloud.google.com/vertex-ai/generative-ai/pricing).

For entity extraction models, consider that the model serving output may be higher as the output is the complete structured data.

## Identify usage of AutoML text APIs

You can determine which of your applications use
AutoML APIs, as well as
which methods they are using. Use this information to help determine whether
these API calls need to be migrated to
Gemini:

- For each of your projects, go to the [APIs \&
  Services Dashboard](https://console.cloud.google.com/apis/dashboard)
  to see a list of
  which products' API the project uses. To learn more, see
  [Monitoring API usage](https://docs.cloud.google.com/apis/docs/monitoring.md.txt).

- If enabled, you can check the audit logs created by
  [AutoML text](https://docs.cloud.google.com/vertex-ai/docs/general/audit-logging)
  as part of [Cloud Audit Logs](https://docs.cloud.google.com/logging/docs/audit.md.txt).

- To see usage of specific AutoML text methods,
  go to the [AutoML text Metrics
  page](https://console.cloud.google.com/apis/api/aiplatform.googleapis.com/metrics).

## Manage changes to IAM roles and permissions

Vertex AI provides the following Identity and Access Management (IAM) roles:

- `aiplatform.admin`
- `aiplatform.user`
- `aiplatform.viewer`

Using Vertex AI datasets is no longer required. Data for fine-tuning in
Gemini can only be stored in Cloud Storage.

For more information on IAM roles, see [Access
control](https://docs.cloud.google.com/vertex-ai/docs/general/access-control).

## What's next

- Read [Introduction to tuning](https://docs.cloud.google.com/vertex-ai/docs/start/migrating-applications)
  in Gemini.

- See [Gemini for AutoML text users](https://docs.cloud.google.com/vertex-ai/docs/start/automl-gemini-comparison.md.txt) for a comparison of Gemini and AutoML text.