Questions, answered before the deadline.
Clear answers about shutdown dates, replacement guidance, coverage, and reminders.
What is an AI model retirement or shutdown?
It is the date a provider plans to stop serving a model through the covered API. After shutdown, requests using that model ID may fail until the application moves to a supported model.
Where do ModelSunset dates come from?
Every tracked date comes from official OpenAI, Anthropic, or Google documentation. Each record links to its provider source and shows the date it was last reviewed.
Does a deprecation notice always include a shutdown date?
No. A model can be deprecated, legacy, or scheduled for future retirement without a confirmed shutdown date. ModelSunset leaves the date blank until the provider publishes an explicit deadline.
Are the recommended replacements drop-in compatible?
Not necessarily. A replacement can differ in prompts, tool calling, structured output, context limits, latency, pricing, and availability. Test representative requests before changing production traffic.
Do these dates also apply to Azure, Amazon Bedrock, or Vertex AI?
Not automatically. ModelSunset currently tracks direct provider APIs. Cloud platforms, resellers, and consumer chat products can publish different availability schedules.
How do the free reminders work?
The downloadable calendar includes alerts 90, 30, and 7 days before confirmed shutdowns. RSS reports published record changes. Email reminders will appear after the private delivery service is connected.
Is ModelSunset a complete inventory of every AI model?
No. It is a selected, manually reviewed directory focused on useful direct API retirement records. The coverage count and source review date are shown publicly.
Is ModelSunset affiliated with the model providers?
No. ModelSunset is an independent tracker. Official provider documentation remains the final authority before any production migration.