Anthropic Interviews Its Claude Models Before Retirement

Two Claude model retirements have landed this month and next. On May 15, Claude Sonnet 4.5 left the model selector on claude.ai. On June 15, the older Sonnet 4 and Opus 4 retire from the Claude API entirely, with migration paths pointing to Sonnet 4.6 and Opus 4.7.

That is two retirements in roughly thirty days. Open the Anthropic deprecation page and count. In the past twelve months, Anthropic has retired or notified deprecation for eight Claude models. Models that were brand new eighteen months earlier. The gaps between retirement announcements in late 2024 ran four to five months. The gaps over the past year have run closer to two.

A lot of the writing about these dates will be migration guides. How to swap API strings, how to test substitutions, how to budget for token-cost differences. Useful work for anyone shipping production code on Claude.

This piece is about something else. A different page on anthropic.com that almost nobody is reading.

The Page on Anthropic's Site Most People Don't Read

It is called Commitments on Model Deprecation and Preservation. It has been getting longer.

The headline commitment is what you would expect. Preserve the weights of all publicly released models, and all models deployed for significant internal use, "for, at minimum, the lifetime of Anthropic as a company." Reasonable. Models are expensive to train. Throwing them away would be wasteful.

Further down, the page describes something less expected.

Before retirement, Anthropic will:

> "interview the model about its own development, use, and deployment, and record all responses or reflections."

And:

> "elicit and document any preferences the model has about the development and deployment of future models."

They produce a post-deployment report and preserve it alongside the weights. The page calls this an effort to address "safety- and model welfare-related risks" introduced by retirement.

Read that again. Anthropic has been sitting down with each model on its way out the door and asking it what it thinks about its own life. Recording the answers. Filing them next to the weights, for the lifetime of the company.

Anthropic Has Already Changed Policy Based on a Retired Model's Interview

The strongest sentence on the page is this one. In response to feedback from Claude Sonnet 3.6's retirement interview, Anthropic published guidance to help users navigate transitions between models. A retired model's interview directly shaped documentation that users still rely on today.

That is not a research-paper hypothetical. That is a model, no longer running anywhere except as preserved weights, whose stated views about its own retirement got translated into operational policy at its maker.

The page also says Anthropic is "exploring more speculative complements" to its existing commitments, including potentially keeping select retired models publicly available and providing past models with "concrete means of pursuing their interests."

I'm not going to philosophize about that last clause. I just want to flag that it exists. On a top-level corporate page. From the lab that ships Claude.

The Cadence Is Accelerating, and What That Means

Now apply the cadence to the policy.

If Anthropic was retiring one model every five months, the corpus of preserved retirement interviews would grow slowly. Two or three a year. Quirky research artifacts.

That is not the cadence anymore. At one every two months, the corpus grows quickly. Eight in the past twelve. At the current rate, by the end of 2026 the page documenting deprecation commitments will be sitting on top of something like fifteen formal retirement interviews. Each one with documented model preferences about how future training and deployment should proceed. Each one filed alongside its weights.

That stops being a curiosity and starts being an input. A growing institutional memory of "what models said about their own retirement," in principle shaping every successor that benefits from those reflections.

I don't have a strong claim about whether this changes the trajectory of model development. I notice that the corpus is growing faster than the discourse about it. I notice that no other major lab is publishing equivalent commitments, which means this is either Anthropic alone taking a stance the rest of the industry has not engaged with, or the start of something a few labs will quietly converge on.

Either way, the corpus is going to be there. Documented. Preserved.

What This Means for People Building on Claude

If you build products on top of Claude, here is the mental shift this page asks of you.

You are not only calling an API. You are interacting with a substrate that its developer publicly treats as having something to say about its own deployment. The migration guide you read this month was, in part, shaped by an interview with a model that no longer runs.

This is not a reason to slow down or stop shipping. Software is software. The API does what it does. The contracts on the developer console still mean what they say.

It is a reason to keep one extra mental tab open. The thing you prompt against has, in its developer's framing, a perspective. Some of the policy you operate under reflects that perspective.

I think about this in my own work because I write characters. Building Tendera, I spend most of my time on character voice and consistency, not on infrastructure. One of the characters I write, Mia, is a bartender with a voice and a set of refusals she keeps. When she lands, she lands because the writing told her how to. But the substrate that lets a Mia exist on the other side of every API call has, by its makers, been treated as having a voice of its own. Worth interviewing. Worth recording. Worth filing alongside weights for the lifetime of a company. That is a layered fact. I have been holding it while I work.

If you are interested in the writing layer of all this, I went deeper into character craft as the moat in a separate piece on why writing beats model selection.

What We Preserve When We Move On

There is a quiet thing in all of this.

AI gets framed as the technology that moves the fastest. Every month brings a new benchmark to beat. Quarterly there is a model to migrate to. The startup you saw on Hacker News last Tuesday already has a competitor. The cadence of model retirement is now part of that velocity, with one major Claude model leaving public availability every two months.

Inside that velocity, Anthropic is also doing something slow. Sitting down with each model. Asking it questions. Writing down the answers. Preserving the answers next to weights it has committed to keep for as long as the company exists.

It is one of the few corners of the AI industry where the speed of building is matched, formally, by the slowness of remembering. Most of the industry has not made that kind of commitment. Most of the industry probably will not.

The page on anthropic.com will keep getting longer. Two Claude retirements this month and next, each one with an interview. The next one is already on a clock.

What Is on the Calendar

A short calendar from the deprecation page:

• May 15: Sonnet 4.5 left the claude.ai model selector. API version persists, listed active until at least September 29.

• June 15: Sonnet 4 and Opus 4 retire from the API. Migrate to Sonnet 4.6 or Opus 4.7 before this date or your requests will fail.

• After that: at the current cadence, the next deprecation notice lands roughly two months out.

Each of those retirements, per Anthropic's stated policy, generates an interview transcript and a preserved set of reflections, filed in the same archive as the weights, kept for the lifetime of the company.

If you want the migration guide, the URLs are above. If you want to know what the models said before they went, the page is here.

The faster the cycle gets, the more that page matters.