Generative AI models maintenance policy

This article describes the model maintenance policy for the Foundation Model APIs pay-per-token and Foundation Model Fine-tuning offerings.

In order to continue supporting the most state-of-the-art models, Databricks might update supported models or retire older models for the Foundation Model APIs pay-per-token and Foundation Model Fine-tuning offerings.

Model retirement policy

The following retirement policy applies only to supported chat and completion models in the Foundation Model APIs pay-per-token and Foundation Model Fine-tuning offerings.

When a model is retired, it is no longer available for use and is removed from the indicated feature offerings. Databricks takes the following steps to notify customers about a model that is set for retirement:

A warning message displays in the model card from the Serving page of your Databricks workspace that indicates that the model is planned for retirement.
A warning message displays in the dropdown menu for Foundation Model Fine-tuning in the Experiments tab that indicates that the model is planned for retirement.
The applicable documentation contains a notice that indicates the model is planned for retirement and the start date it will no longer be supported.

After users are notified about the upcoming model retirement, Databricks will retire the model in three months. During this three-month period, customers can either:

Choose to migrate to a provisioned throughput endpoint to continue using the model past its end-of-life date
Migrate existing workflows to use recommended replacement models.

On the retirement date, the model is removed from the product, and applicable documentation is updated to recommend using a replacement model.

See Retired models for a list of currently retired models and planned retirement dates.

Model updates

Databricks might ship incremental updates to pay-per-token models to deliver optimizations. When a model is updated, the endpoint URL remains the same, but the model ID in the response object changes to reflect the date of the update. For example, if an update is shipped to meta-llama/Meta-Llama-3.1-405B on 3/4/2024, the model name in the response object updates to meta-llama/Meta-Llama-3.1-405B-030424. Databricks maintains a version history of the updates that you can refer to.

Retired models

The following sections summarize current and upcoming model retirements for the Foundation Model APIs pay-per-token and Foundation Model Fine-tuning offerings.

Foundation Model Fine-tuning retirements

The following table shows retired model families, their retirement dates, and recommended replacement model families to use for Foundation Model Fine-tuning workloads. Databricks recommends that you migrate your applications to use replacement models before the indicated retirement date.

Model family	Retirement date	Recommended replacement model family
Meta-Llama-3	January 7, 2025	Meta-Llama-3.1
Meta-Llama-2	January 7, 2025	Meta-Llama-3.1
Code Llama	January 7, 2025	Meta-Llama-3.1

Foundation Model APIs pay-per-token retirements

The following table shows model retirements, their retirement dates, and recommended replacement models to use for Foundation Model APIs pay-per-token serving workloads. Databricks recommends that you migrate your applications to use replacement models before the indicated retirement date.

Important

On December 11, 2024, Meta-Llama-3.3-70B-Instruct replaced support for Meta-Llama-3.1-70B-Instruct in Foundation Model APIs pay-per-token endpoints.

Model	Retirement date	Recommended replacement model
Meta-Llama-3.1-70B-Instruct	December 11, 2024	Meta-Llama-3.3-70B-Instruct
Meta-Llama-3-70B-Instruct	July 23, 2024	Meta-Llama-3.1-70B-Instruct
Meta-Llama-2-70B-Chat	October 30, 2024	Meta-Llama-3.1-70B-Instruct
MPT 7B Instruct	August 30, 2024	Mixtral-8x7B
MPT 30B Instruct	August 30, 2024	Mixtral-8x7B

If you require long-term support for a specific model version, Databricks recommends using Foundation Model APIs provisioned throughput for your serving workloads.