AI Model Deprecation & Lifecycle Tracker
Is the model you're calling about to be turned off? Search any OpenAI, Anthropic, Google, or Azure model ID to see whether it's active, deprecated, or retired — with the official shutdown date, a live countdown, and the recommended replacement. No signup, sources cited below.
How it works
This is a reference, not a calculator: the lifecycle dates themselves are editorial. Each model's deprecation date, shutdown (retirement) date, status, and recommended replacement are transcribed verbatim from the provider's own deprecations or model-retirements page, and every row links back to the exact source. Nothing is inferred — where a provider has announced a deprecation but not a firm shutdown date, the tool shows “—” rather than a guessed value.
Models move through four lifecycle states, which the tracker re-derives on every load:
- Active — current and recommended; no deprecation announced.
- Legacy — still callable but superseded by a newer version, with no shutdown date published yet.
- Deprecated — officially scheduled for shutdown:
deprecationDate ≤ today < shutdownDate. The model still works during this notice window. - Retired — the shutdown date has passed (
today ≥ shutdownDate), so API calls now return errors. The stored status is authoritative, but the tool re-derives Deprecated → Retired the instant a shutdown date passes, so a past date is never shown as still “deprecated.”
The only computed value is the countdown, and it is deterministic calendar arithmetic done entirely at UTC midnight to avoid any timezone off-by-one:
- daysUntilShutdown = epochDay(shutdownDate) − epochDay(today)
- Positive → “shuts down in N days.”
- Zero → “shuts down today.”
- Negative → Retired; “shut down |N| days ago.”
That subtraction is implemented twice — once via integer day indices from Date.UTCand once via millisecond parsing of an explicit UTC timestamp — and the two are required to agree, so an arithmetic regression would be caught rather than shipped. The summary strip's “next shutdown” highlight is simply the soonest shutdown date strictly after today across the whole dataset. Source for the policy that deprecation precedes shutdown by a notice window: OpenAI's deprecations page and Anthropic's model-deprecations policy, which both state models stay callable between the announcement and the shutdown.
Worked examples
Reference point for all three: 13 June 2026 (the 2026-06-13 snapshot date). The live tool recomputes against the actual current date.
Frequently asked questions
Sources & references
Each model row links to the specific provider page it was transcribed from. The canonical deprecation / retirement pages:
- OpenAI — Deprecations (model deprecation & shutdown dates)
- Anthropic — Model Deprecations (retired Claude models & policy)
- Google — Gemini API model versions & discontinuations
- Microsoft — Azure OpenAI Service model retirements
Dates were last cross-checked against these sources on 2026-06-13. This is a cited snapshot, not a live feed — provider pages change, so the verified date above always tells you how current the data is.
Related tools
Comments & feedback
Spotted a bug or want an improvement? Tell us — our team reviews every comment, and good ideas get built. Comments are public and anonymous.
Found an out-of-date row, a model we're missing, or an edge case?
Email me at [email protected] — most fixes ship within 24 hours.