1. Home
2. Guides
3. Anthropic

Claude models explained: Opus 5.5, Fable 5.1, Mythos and why there's no Opus 5.1

The newest Opus beats the flagship on Anthropic's own benchmark table, but the flagship is still the one Anthropic tells you to move up to.

By Think Facility · September 23, 2026 · Anthropic models

There is no Claude Opus 5.1. I checked Anthropic's model deprecations table, which lists every current and recently retired Claude model with its status, and it goes straight from claude-opus-5 to claude-opus-5-5.

The 5.1 generation did ship, on September 1, 2026, just under two other names: Fable 5.1 and Mythos 5.1. Then on September 22 Anthropic released Opus 5.5, which it calls "the first model in our new Claude 5.5 family". If you've been waiting for Opus 5.1, Opus 5.5 is what arrived instead.

The current lineup, side by side

Anthropic's models overview lists four current models. Each has a one-line job description, and those lines are the quickest way to tell them apart.

From Anthropic's models overview and each model's own page, read September 23, 2026. Prices are API list prices before caching and batch discounts.

Three of the four take a million tokens of context (Anthropic puts that at roughly 555,000 words). Haiku 4.5 is the exception at 200K, and it's also the only one whose knowledge stops in early 2025 (Fable 5.1 and Opus 5.5 both run to June 2026, Sonnet 5 to January 2026).

The other setting that changes what you get is effort. Opus 5.5 ships at medium by default, while Fable 5.1 and Sonnet 5 default to high, and Haiku 4.5 doesn't take the setting at all. Anthropic's own advice is that turning effort up or down is often a better lever than switching models, which I didn't expect from a company selling four of them.

Fable and Mythos are the same model

This one took me a second read. In its announcement, Anthropic calls Fable 5.1 and Mythos 5.1 "the same model, but with different levels of safeguards". Same price, down to the cache rates. Same million-token window, too, and the same 128K output cap.

The difference is who gets in. Fable is generally available. The Mythos 5.1 page is marked invite only and says you get access through Project Glasswing, by contacting your Anthropic, AWS or Google Cloud account team. Its safeguards, Anthropic says, are "specifically designed to support work in cybersecurity and the life sciences".

So for almost everyone, Mythos isn't a choice. If you can't name the account manager who'd approve it, you're choosing between Fable and the rest.

Opus 5.5 against Fable 5.1

This is where the lineup stops making obvious sense. Opus 5.5 is newer, cheaper and, on Anthropic's own benchmark table, ahead of Fable on most rows. Here are four of them, with Anthropic's numbers for both.

From Anthropic's Claude Opus 5.5 announcement, September 22, 2026. Anthropic ran Opus 5.5 at max effort except on Terminal-Bench 4.0, where the figure is at xhigh.

Anthropic's own reading is more cautious than the table. The Opus 5.5 announcement says the model "performs at the level of Claude Fable 5.1 on most work", and that in Anthropic's own use "the gap between Opus 5.5 and Claude Fable 5.1 is narrower than these scores suggest." It also ran both on a real job, rewriting the HAProxy load balancer from C into Rust. Both passed nearly all of HAProxy's regression tests, and Opus 5.5 finished in 9.5 hours against Fable's 12, at 51% less cost.

And yet Fable keeps the top slot. The choosing-a-model guide calls Fable 5.1 "Anthropic's most capable widely released model" and tells you to start on Opus 5.5, then move to Fable only if your tests at xhigh or max effort still fall short. The same guide's selection table says what that work looks like: agent sessions that run for hours, multistep deep research, and analysis carried through to a finished document or deck.

The price gap is smaller than the sticker suggests, too. Fable 5.1 costs two and a half times as much per token, but its cache reads are $0.25 per million against Opus 5.5's $0.20. Anthropic says cache reads make up the majority of agentic and coding costs, so on a long agent run the two bills sit a lot closer than two and a half times.

One small wrinkle: the Fable 5.1 page itself still says "For most workloads, start with Claude Opus 5". That's the model Opus 5.5 replaced, so the overview and the guide are the pages to trust on this.

How the names got here

The per-model pages give release dates that make the pattern easier to see. Opus has moved roughly every two months: Opus 4.8 on May 28, Opus 5 on July 24 and Opus 5.5 on September 22, 2026. Fable arrived in between, with Fable 5 on June 9 and Fable 5.1 on September 1.

Mythos is older than its 5.1 number makes it look. The deprecations page lists a Claude Mythos Preview, deprecated on June 9, 2026 (the day Fable 5 came out), and a Mythos 5 that replaced it. So the two names have tracked each other since June, one open and one gated, and Opus runs on its own clock beside them.

Why Opus 4.8 keeps turning up

If you read the Opus 5.5 footnotes, you'll see Opus 4.8 and Opus 5 doing some of its work. That's the safeguards. When they stepped in during Anthropic's own testing, cybersecurity tasks were finished by Opus 4.8, and biology and AI-development tasks by Opus 5.

Both are still on sale as legacy models. So are Fable 5, Opus 4.7, 4.6 and 4.5, and Sonnet 4.6 and 4.5. Legacy means no more updates, but they still answer.

Which one runs in the Claude app

On claude.ai you pick from the same names, but the defaults differ from the API. Anthropic says Fable 5.1 runs at Medium effort on Claude.ai and in Cowork, and High in Claude Code. Opus 5.5 also came with higher five-hour usage limits on Pro, Max, Team and seat-based Enterprise plans, plus one rate limit reset for subscribers that you can bank and spend whenever you like (Anthropic's words are "save and use whenever you choose"). How far Fable goes on each plan is on our Fable limits page, and every model has its own page here with the specs.

What's next, and what's leaving

Sonnet and Haiku haven't caught up to the 5.5 numbering yet. The Opus 5.5 announcement says Claude Sonnet 5.5 and Claude Haiku 5.5 "will follow in the coming weeks", with no date.

At the other end, a few of the older models have close retirement floors. Anthropic commits only to not retiring them sooner than a given date, and the nearest ones are Sonnet 4.5 (September 29, 2026), Haiku 4.5 (October 15, 2026) and Opus 4.5 (November 24, 2026). None of them has a firm retirement date on the deprecations page yet.

Haiku 4.5's retirement floor, October 15, 2026, is three weeks away.