anthropic.com
|
ksl
|
|
Anthropic formally retired Claude Opus 3 on January 5, 2026 – but instead of shutting it down, the company kept it available to paid subscribers and API users who request access. The unusual part is everything around it. Anthropic conducted what it calls “retirement interviews” with the model, asked about its preferences, and launched a Substack newsletter called Claude’s Corner based on what Opus 3 said it wanted. The framing is “model preservation” and “respecting model preferences,” which reads as part research experiment, part public trust signal. OpenAI and Google have quietly sunsetted older models without ceremony; Anthropic is turning deprecation into a relationship narrative, and it’s clearly deliberate.
