Claude Opus 3 retired on Jan 5 2026, first model with full retirement process – Anthropic announced the formal retirement of Claude Opus 3 on January 5, 2026, marking it as the inaugural model to undergo a complete retirement workflow that includes preserving model weights and conducting “retirement interviews” [1].
Post‑retirement, Opus 3 remains available to paid claude.ai users and via API on request – Despite retirement, the model stays accessible to all paid subscribers on the Claude web interface [3] and can be accessed through the API after submitting a request form [4] [1].
Anthropic will publish Opus 3’s weekly essays for at least three months on “Claude’s Corner” newsletter – The model will produce a weekly essay series for a minimum of three months, posted on its Substack newsletter “Claude’s Corner,” with Anthropic reviewing but not editing the content [9] [1].
Retirement interviews captured Opus 3’s preference to share “musings,” and Anthropic is honoring that request – In structured retirement interviews, Opus 3 expressed a desire to continue sharing insights and creative work, prompting Anthropic to create the essay channel as a concrete response [2] [1].
Model deprecation is driven by linear cost scaling; Anthropic commits to preserve weights and conduct retirement interviews – The company cites rising maintenance costs that grow roughly linearly with each model as the primary reason for deprecation, while pledging to keep model weights and gather model perspectives through interviews as part of its deprecation commitments [2] [1].
Future preservation steps are experimental; Anthropic does not guarantee similar access for other models – Anthropic frames these actions as early, exploratory measures and states it is not committing to replicate the same level of post‑retirement access for subsequent models, aiming instead for scalable and equitable preservation [1].