Model versions and lifecycle

This guide defines key terms related to the lifecycle stages of Gemini and embedding models, and provides information on important dates, recommended upgrades, and migration paths.

This guide covers the following topics:

Key Terms

  • Stable model: A publicly released version of a model that is available and supported for production use. A stable model version typically has a retirement date, which is the last day that the model is available. After this date, the model is no longer accessible or supported by Google.

    • Latest stable model: The most recent version within a model family. Google recommends this version for new and active projects. It should also be your target when you migrate from earlier versions. See Latest stable models.
    • Legacy stable model: A model version that a newer version has superseded. Although legacy stable models are still supported, you should migrate to the latest model to get the latest features and improvements. Google might restrict access to legacy stable models for new projects. See Legacy stable models.

  • Retired model: A model version that is past its retirement date and is permanently deactivated. Google no longer supports retired models, and they are inaccessible. API requests that reference a retired model ID typically return a 404 Not Found error. See Retired models.

  • Recommended upgrade: The latest stable model that you should upgrade to. Latest stable models typically offer better performance and more capabilities than legacy stable models. For specific recommendations, see the upgrade paths in the Legacy stable models and Retired models sections.

Latest stable models

The following table lists the latest stable models:

Model ID Release date Retirement date Details
gemini-2.5-pro June 17, 2025 June 17, 2026
gemini-2.5-flash June 17, 2025 June 17, 2026
gemini-2.5-flash-lite July 22, 2025 July 22, 2026
gemini-2.0-flash-001 February 5, 2025 February 5, 2026 Gemini 2.0: Flash, Flash-Lite and Pro - Google Developers Blog
gemini-2.0-flash-lite-001 February 25, 2025 February 25, 2026 Gemini 2.0: Flash, Flash-Lite and Pro - Google Developers Blog
gemini-embedding-001 May 20, 2025 No retirement date announced
text-embedding-005 November 18, 2024 No retirement date announced
text-multilingual-embedding-002 May 14, 2024 No retirement date announced
multimodalembedding@001 February 12, 2024 No retirement date announced

Legacy stable models

The following table lists legacy stable models:

Model ID Release date Retirement date Recommended upgrade
gemini-1.5-pro-002* September 24, 2024 September 24, 2025 gemini-2.0-flash
gemini-1.5-flash-002* September 24, 2024 September 24, 2025 gemini-2.0-flash-lite
text-embedding-004 May 14, 2024 November 18, 2025 gemini-embedding-001

*Restricted for new projects.

Migrate to a latest stable model

To learn how to migrate to a latest stable model, see Migrate your application to Gemini 2 with the Gemini API in Vertex AI. This guide provides migration steps to help you minimize risks and use new models effectively.

If you need to update your application to resolve errors caused by a model reaching its retirement date, do the following:

  1. Update your application to point to the recommended upgrade.
  2. Test all mission-critical features to make sure that they work as expected.
  3. Deploy the updates.

Gemini auto-updated aliases

You can specify a model using either a pinned version ID (for example, gemini-1.5-pro-002) or an auto-updated alias (for example, gemini-1.5-pro). An auto-updated alias for a Gemini model always points to the latest stable version of that model. When a new stable version is released, the alias automatically points to the new version.

The following table compares these two options.

Option Description Pros Use Case
Pinned version (for example, gemini-1.5-pro-002) Uses a specific, unchanging model version. Predictable behavior and stable outputs, as the underlying model does not change. Production environments where consistency and reproducibility are critical.
Auto-updated alias (for example, gemini-1.5-pro) Automatically points to the latest stable version of a model. Access to the latest features and improvements without manual code changes. Development or non-critical applications where leveraging the newest model capabilities is prioritized over strict version consistency.

The following table shows the auto-updated aliases for Gemini models and the latest stable models that they point to.

Auto-updated alias Stable version reference
gemini-2.0-flash-lite gemini-2.0-flash-lite-001
gemini-2.0-flash gemini-2.0-flash-001
gemini-1.5-pro gemini-1.5-pro-002
gemini-1.5-flash gemini-1.5-flash-002

Retired models

View retired models

Model ID Release date Retirement date Recommended upgrade
gemini-1.5-pro-001 May 24, 2024 May 24, 2025 gemini-2.0-flash
gemini-1.5-flash-001 May 24, 2024 May 24, 2025 gemini-2.0-flash-lite
textembedding-gecko@003* December 12, 2023 May 24, 2025 gemini-embedding-001
textembedding-gecko-multilingual@001 November 2, 2023 May 24, 2025 gemini-embedding-001
gemini-1.0-pro-001 February 15, 2024 April 21, 2025 gemini-2.0-flash
gemini-1.0-pro-002 April 9, 2024 April 21, 2025 gemini-2.0-flash
gemini-1.0-pro-vision-001 February 15, 2024 April 21, 2025 gemini-2.0-flash
text-bison May 2023 April 21, 2025 gemini-2.0-flash-lite
chat-bison May 2023 April 21, 2025 gemini-2.0-flash-lite
code-gecko May 2023 April 21, 2025 gemini-2.0-flash-lite
textembedding-gecko@002 November 2, 2023 April 21, 2025 gemini-embedding-001
textembedding-gecko@001 June 7, 2023 April 21, 2025 gemini-embedding-001

What's next