Anthropic retired Claude Sonnet 4 and Claude Opus 4 on 15 June. Developers were warned on 14 April, so this was planned, but the effect is blunt: any tool that calls one of those models by name now fails instead of running.
More is coming. Claude Opus 4.1 is deprecated with a 60-day window before it retires on 5 August. The fast mode for Opus 4.7 ends on 24 July, after which a request asking for fast speed returns an error.
Why This Catches Businesses Out
When a developer builds an AI feature, they often pin it to the exact model that worked during testing. That is sensible at the time. The problem is that models get retired on a schedule the business never sees. The automation runs fine for months, then one morning it returns errors and nobody knows why, because nothing in the business changed. The model underneath did.
What to Check
If you run any AI tool, chatbot, or automation, ask whoever built it one question: which model does it call, and is that model still supported? If the answer is Sonnet 4, Opus 4, or a fast mode on an older Opus, you have a fix to schedule before the dates above.
The fix is usually small. Point the tool at a current model and test the output. But small and unplanned still means downtime if you find out by the feature breaking in front of a customer. Better to check now, while it is a calendar task and not an emergency.