Anthropic gives API users a 60-day retirement notice. What about Claude.ai users?
retained text
Haiku 5.5 shipped this week. The [claude.ai](http://claude.ai) picker now shows four models up front and nine more under "More models," going all the way back to Opus 3. With more launches coming, this seems like a good moment for a plain process question. **Anthropic's published deprecation policy includes a 60-day advance-notice window for API model retirements. Removals from the** [**claude.ai**](http://claude.ai) **model picker currently have no equivalent notice mechanism.** This isn't hypothetical. Opus 4.5 disappeared from the [claude.ai](http://claude.ai) model picker around the launch of Opus 4.7 in April 2026. There was no announcement, no retirement interview, and no user notification. Users discovered it by opening the picker and seeing it gone. One detail matters here: Opus 4.5 didn't stop working. Existing conversations with it still function. It was removed from selection, not from existence. So this wasn't a capacity decision forced by retirement, but a picker change made silently around a new launch. **Why this matters** Users who've built workflows, writing practices, or long-term habits around a specific model can lose it from the app without warning. That's a structural gap, and it's fixable. Anthropic itself has acknowledged that model retirements come with costs, particularly for users who value specific model characteristics. In its November 2025 commitments, it wrote: "Each Claude model has a unique character, and some users find specific models especially useful or compelling, even when new models are more capable." When I asked support about this, the reply confirmed the gap: the 60-day policy applies to API deprecations, and the documentation doesn't say whether it covers [claude.ai](http://claude.ai) picker changes. Proposals for legacy access, they said, are a product-policy decision beyond what support can handle. **The question** If a model is going to be removed from the [claude.ai](http://claude.ai) picker, should users get the same advance notice that API deprecations already have? And if a model has developed distinctive behavioral value for a group of users, should there be some path to continued, limited access — as was done for Opus 3? I'm not arguing that every old model should be kept forever. I'm arguing that removals shouldn't be silent. "The new model is more capable" and "users still need the old one" can both be true. Anthropic's own commitments say this. If this affects you, describing your own experience in the comments is more useful than upvotes. Specific use cases carry weight.
document fetched 2026-10-09 00:22 · canonical https://reddit.com/r/ClaudeAI/comments/1x15ffd/anthropic_gives_api_users_a_60day_retirement