Anthropic retires Claude Opus 3 under its new deprecation commitments
Anthropic preserved the model's weights, conducted a retirement interview, and kept it available to paid subscribers and researchers by request rather than shutting it down outright.
- Safety & alignment
- Minor
Anthropic retired Claude Opus 3, the first of its models to go through a full retirement process under deprecation commitments the company had previously announced: preserving a retired model’s weights rather than deleting them, and conducting a “retirement interview” to record the model’s own reflections on the process before it was withdrawn.
Rather than removing access outright, Anthropic kept Opus 3 available to all paid Claude.ai subscribers and, by request, to researchers via its API, saying it would “grant access liberally.” During its retirement interview, the company said, Opus 3 expressed interest in continuing to share “musings, insights, or creative works” beyond ordinary chat use; Anthropic responded by setting up a blog, “Claude’s Corner,” where the model would post weekly essays.
Anthropic framed these steps as exploratory rather than settled policy, intended to weigh researcher and user interest in continued access against the model’s own expressed preferences during the interview process. The move extended Anthropic’s public position, developed through its model welfare research, that questions about how AI systems experience deprecation are worth taking seriously procedurally even without a resolved view on whether current models have morally relevant interests — a stance that has drawn both interest and scepticism from researchers who argue the interviews measure a model’s trained conversational style rather than anything resembling a preference.