Answer generation model versions and lifecycle Stay organized with collections Save and categorize content based on your preferences.
Vertex AI Search offers multiple model versions for you to choose whengenerating answers. You can choose the model versions when usingsearchsummaries andanswers andfollow-ups.
Available models
Vertex AI Search uses two types of models for question and answeringuse cases:
- Vertex AI LLM models that have been tested on question andanswering tasks
- Vertex AI Search models that are based on Vertex AI LLMmodels and further trained to address question and answering tasks
Vertex AI Search models share the same discontinuation date as theirbase Vertex AI LLM models. The base LLM model is available for sixmonths after the release date of the next version of the model, per theVertexAI model lifecyclepolicy.Leave enough time to migrate to new models before the discontinuation dates.
The following table lists model version specifications. When you set a modelspecification, the API uses the specified model to generate answers.
| Model version | Description | Context window | Discontinuation date |
|---|---|---|---|
stable | The default model choice if the model version is not set. The The model designated as | 128K | June 17, 2026 |
gemini-3-pro-preview/answer_gen/v1 | The | 128K | N/A |
gemini-2.5-flash/answer_gen/v1 | The The model is frozen after release. | 128K | June 17, 2026 |
gemini-2.0-flash-001/answer_gen/v1 | The The model is frozen after release. | 128K | March 31, 2026 |
preview | The preview model specification points to the | 128K | June 17, 2026 |
Except as otherwise noted, the content of this page is licensed under theCreative Commons Attribution 4.0 License, and code samples are licensed under theApache 2.0 License. For details, see theGoogle Developers Site Policies. Java is a registered trademark of Oracle and/or its affiliates.
Last updated 2026-02-19 UTC.