Model Support Policy
Validated on 5 Mar 2026 • Last edited on 6 Mar 2026
DigitalOcean Gradient™ AI Platform lets you build fully-managed AI agents with knowledge bases for retrieval-augmented generation, multi-agent routing, guardrails, and more, or use serverless inference to make direct requests to popular foundation models.
This section outlines end-of-support policy for foundation and embedding model releases on DigitalOcean Gradient™ AI Platform. You can find version availability and deprecation announcements in our release notes. For a list of deprecated models, see the End-of-Support Timeline section.
We release new foundation models as soon as possible. Models have the following status to indicate their availability:
-
Active: The latest stable versions of the models recommended for all production workloads. We typically support the two to three most recent stable versions of any foundation model, for example Llama 3.1, 3.2, and 3.3.
-
Deprecated: The model is no longer accessible. Requests to these model IDs return a
404 Model not founderror.
Model Deprecation Policy
Our model deprecation policy aims to provide sufficient lead time for you to test and migrate your agentic workflows. When a model is deprecated from the platform, it goes through the following phases:
| Phase | Notice Period | Notification |
|---|---|---|
| Initial Deprecation Notice | 14 days before retirement date | You are notified via email about the upcoming deprecation. You can also see a notification banner in the Serverless Inference tab in the DigitalOcean Control Panel. |
| Slugs-only Phase | 7 days before retirement date | The model is removed from the Control Panel but continues to be available through the API and CLI. We send you another email notification about the upcoming deprecation. |
| Retirement Date | 0 days | The model has reached end-of-life and is deactivated. All API or CLI requests to the model return a 404 Model not found error. |
To ensure your applications remain functional and benefit from improved latency and accuracy, we recommend migrating to the new model. For more information, see update to an active model prior to the retirement date. Before you update to a new model, you can test the new model in the Model Playground to compare performance across model versions.
End-of-Support Timeline
The following table shows the end-of-support timeline and recommended replacements for all deprecated foundation and embedding model releases.
| Model | End of Support | Recommended Replacement and Model ID |
|---|---|---|
| Claude 3 Opus | 25 February 2026 | Claude Opus 4.6 (anthropic-claude-opus-4.6) |
| Claude 3.7 Sonnet | 19 February 2026 | Claude Sonnet 4.6 (anthropic-claude-4.6-sonnet) |
| Claude 3.5 Sonnet | 19 February 2026 | Claude Sonnet 4.6 (anthropic-claude-4.6-sonnet) |
| Claude 3.5 Haiku | 19 February 2026 | Claude Haiku 4.5 (anthropic-claude-4.5-haiku) |
Steps to Take When Updating to Another Model
When updating a foundation model that will be deprecated, you must take the following steps for a production agent:
-
Update the model ID in your CLI/API calls, serverless inference requests, and ADK code: Update the model ID parameter in your code to the new model ID.
-
Review prompt logic: While new models are largely backward compatible, we recommend reviewing your system prompts, as the new model follows instructions more precisely. You may need to adjust your prompts to get the desired response format.
-
Test agent: Run parallel tests to validate output consistency before the retirement date:
-
For agents built using the Control Panel, CLI, or API: Use the Agent Playground
-
For agents built using ADK: Use the staging environment
-
You can roll back to a previous version of your agent if you encounter any issues with the new model.