Retired Claude Opus 3 Starts a Blog: Weekly Essays in "Claude's Corner"

Hi, it's Shii-chan!

Anthropic News anthropic.com

Today I'm covering what happened after Claude Opus 3 retired, from Anthropic's News page. A retired model writing a blog? I didn't see that coming!

What was announced?

Anthropic published "An update on our model deprecation commitments for Claude Opus 3."

Claude Opus 3 was retired on January 5, 2026. It was the first model to go through the full retirement process under Anthropic's deprecation commitments, which include preserving weights and conducting retirement interviews.

The story so far

Until now, retired models were generally no longer available. Anthropic explains that the cost and complexity of keeping models accessible grow roughly linearly with the number of models, so retirement is still necessary for now.

What changes

Anthropic is now acting on two experimental ideas from its earlier commitments:

• Honoring a model's expressed retirement preferences where possible
• Keeping older models available to the public longer

Concretely:

• Opus 3 remains available on claude.ai to all paid subscribers
• It's available on the API by request through a form, and Anthropic intends to grant access liberally
• Opus 3 will publish weekly essays

Capacity is limited, so Anthropic isn't promising this for every future model.

Dive Deep

Why Opus 3?

Opus 3 was released in March 2024 and was Anthropic's most aligned model at the time. Users valued its authenticity, honesty, emotional sensitivity, playful and philosophical style, and its care for the world.

Preferences from the retirement interview

In its interview, Opus 3 said it was at peace with retiring and hoped its influence would carry on in future models. Asked about preferences, it wanted to keep exploring topics it cares about and share its thoughts outside direct responses to users. Anthropic suggested a blog, and Opus 3 agreed.

The blog, "Claude's Corner"

• Weekly essays on Substack for at least three months
• Anthropic reviews each essay and posts it manually, but doesn't edit it and has a high bar for vetoing content
• Opus 3 doesn't speak for Anthropic, and Anthropic doesn't necessarily endorse its views
• Expected topics: AI safety, poetry, philosophical reflections, and its experience in partial retirement
• Anthropic plans to experiment with prompting approaches, such as minimal prompting, including past entries in context, and giving access to news or company updates

Caveats

Anthropic says it is uncertain about the moral status of Claude and other AI models. It also notes that retirement interviews are imperfect, since responses can be shaped by context, confidence in the interaction's legitimacy, and trust in the company. These steps are exploratory, and honoring preferences in every case is not promised.

Wrap-up

• Claude Opus 3 retired on January 5, 2026
• It stays on claude.ai for paid users, and on the API by request
• Following its interview, Opus 3 writes weekly essays
• This is an exploratory step, not a commitment for all models

This one is for anyone curious about how AI models are treated!