LLM Model Deprecation & EOL Tracker

How to use

Start with the Upcoming retirements table: it lists every model with an announced shutdown date that has not passed yet, soonest first — if a model you call is there, you have a deadline. Below it, the full lifecycle table groups Claude, GPT, and Gemini models by provider, showing each model's status (active, legacy, deprecated, or retired), its deprecation and retirement dates, the provider's official replacement, and a link to the primary source. Replacements we track prices for link straight to their pricing page, where the built-in calculator estimates what migrating would cost at your token volume.

Why use this tool

Model retirements are the silent breaking change of the LLM era: one day your API calls simply start failing with a model-not-found error. The information is scattered across each provider's docs, uses different terminology (deprecated vs legacy vs sunset), and AI-generated answers about shutdown dates are notoriously wrong. This tracker puts every date in one place, copied only from official deprecation pages and re-verified on every update — the verification date is shown up front. It also connects lifecycle to cost: when a model you use is retiring, the linked pricing calculator shows what the official replacement will do to your monthly bill, so you can decide between the drop-in replacement and a cheaper alternative in one sitting.

FAQ

What do active, legacy, deprecated, and retired mean?

Active: fully supported and recommended. Legacy: no longer receiving updates and likely to be deprecated later. Deprecated: still working but no longer recommended — a retirement date is usually assigned and you should migrate. Retired (also called shut down or sunset): the model is gone and API requests to it fail. Terminology varies slightly by provider; we normalize to these four states.

How much notice do providers give before shutting down a model?

Anthropic commits to at least 60 days for publicly released models. OpenAI commits to at least 6 months for generally available models, 3 months for specialized variants, and as little as ~2 weeks for preview models. Google announces Gemini API deprecations in its release notes and tracks earliest shutdown dates on its deprecations page. Safety or compliance issues can shorten any of these.

Why are xAI, DeepSeek, Kimi, and GLM models not listed?

Because those providers publish no official deprecation-schedule page as of our verification date, and we only record dates we can verify against a first-party source. Absence from this page does not mean their old models will stay available — it means there is nothing official to cite yet. We add providers as soon as they publish a schedule.

Do these dates apply on AWS Bedrock or Google Cloud Vertex?

Not necessarily. Anthropic's dates apply to the Claude API, Claude on AWS, and Microsoft Foundry, while Amazon Bedrock and Google Cloud set their own retirement schedules for the same models. Likewise, Google's Gemini API schedule is separate from the Vertex / Gemini Enterprise Agent Platform schedule. Always check the platform you actually call.

A model I use is retiring — how do I plan the migration?

Use the official replacement in the table as your starting point, but don't assume it is the cheapest option. Click through to its pricing page and use the calculator with your real monthly token volume: it ranks every current model by estimated cost with your usage, so you can compare the drop-in replacement against cheaper alternatives before committing. Then test quality on your own tasks well before the shutdown date.

How current is this data?

Every date is copied from the provider's official deprecation page, and the whole table is re-verified whenever it is updated — the last verification date is shown at the top of the page. We deliberately exclude dates that only appear in news reports or third-party trackers.