Model Lifecycle: Models Launch, Change, and Get Retired

Models aren't forever

AI models have a lifecycle: launch, updates, eventual deprecation, and shutdown. This applies across all major providers. If you build an application hard-wired to a single model, you risk it breaking down the line.

Example: Fable 5

Anthropic launched Claude Fable 5 on 2026-06-09. Just three days later, on 2026-06-12, the US government imposed export controls after an Amazon researcher found a jailbreak that let the model identify software vulnerabilities and demonstrate exploits. The controls were lifted on 2026-06-30, and since 2026-07-01 Fable 5 has been available worldwide again, with a new safety classifier. Requests the classifier blocks are now automatically routed to Opus 4.8. Anthropic itself considers the jailbreak a borderline case - weaker models produce similar results too.

Regular deprecations happen too

Beyond special situations like this, providers also retire models through ordinary deprecation cycles as newer versions ship. Anthropic publishes official deprecation information showing when each model reaches end of life.

What this means for you

Build applications to be as model-agnostic as possible: abstract the model call so you can swap models without rewriting your whole codebase. Plan for fallbacks - a second model or a second provider - in case your primary model becomes unavailable on short notice.