Model deprecations and retirement dates

The rows are Salyro's model catalogue, the models Salyro serves and no others. Each model's lifecycle is joined from its provider's inventory snapshot: what the provider's sources state, recorded with its source and the day it was checked.

• Active: the provider's page states that the model is active. Anthropic's gives even an active model a tentative retirement date, before which it will not be retired.
• Deprecated: the provider has announced a date to stop serving the model and has announced it will serve the model until then. It is not retired.
• Retired: the provider says the model is no longer served.
• Unknown: the provider's sources list the model without a lifecycle, or could not be read. That is never taken to mean active.

The date is always the earliest the provider gives. For Gemini 2.5 Flash-Lite, that is the Gemini API's 2026-10-16, not the later 2026-10-20 of Google's Vertex AI surface. The table was recorded on the day in its caption. Once the page loads, a date that has passed since, by your device's date, is marked as passed, and the provider may already have acted on it: Google's API has been reported answering some Gemini 2.5 requests with "no longer available" before its date.

The source column says how each provider's facts were obtained. Anthropic's page was read by the build itself. OpenAI's model list is its Python SDK's. Google's and xAI's pages could not be reached from the build, so their facts were verified by a review done outside it.

Search by model or provider, filter by provider or state, sort by any column, and download the rows you see as CSV. Each row has its own link: select the model's name.