Claude Fable 5 Explained: What It Is and Why It Matters

On June 9, 2026, Anthropic released Claude Fable 5, and it's a bigger deal than the usual point update. This is the first publicly available model in what Anthropic calls its 'Mythos-class' tier, which it describes as the most capable Claude it has ever made generally available. If you've been hearing the name pass around and you're not sure what's actually new, this is the plain explainer.

Quick note before we go further: ShadowProof is independent and we're not affiliated with Anthropic in any way. We build a US mobile IP app for creators, and we write about tools our readers ask about. This piece is just us reading the release and telling you what it says, with no spin.

What Claude Fable 5 actually is

Claude Fable 5 is Anthropic's new flagship model. The headline claim is that it's state of the art on most of the benchmarks Anthropic ran, across a few different areas:

• Software engineering and coding. It posted the top score on Cognition's FrontierCode at medium effort.
• Finance. It came out on top on Hebbia's Finance Benchmark.
• Vision. Anthropic says it can rebuild web apps from screenshots and read numbers off scientific figures.
• Long context. It can hold focus across millions of tokens, and memory improvements give it roughly 3x better performance on complex games.

That last point about long context is the one I'd pay attention to. Plenty of models claim a big context window, but holding actual focus across millions of tokens is a different skill from just accepting them as input. The memory improvements are what drive that 3x figure on the harder, multi-step tasks.

One real example Anthropic shared makes it concrete. Stripe used Fable 5 to finish a 50 million line Ruby migration in a single day. Anthropic says that kind of work would normally take a team about two months. That's the type of long-horizon, many-step job the model is built for.

How it's different from Opus 4.8

Most people will compare Fable 5 to Opus 4.8, since that's the model a lot of teams are running today. Anthropic's claim is straightforward: Fable 5 beats Opus on nearly all of its benchmarks, and the gap is widest on long-horizon complex tasks, finance, and coding.

The catch is price. Fable 5 costs double what Opus 4.8 costs. So the real question isn't 'which is better' so much as 'is the harder task worth paying twice as much for.' For a quick chat or a simple script, probably not. For something like a sprawling code migration or a deep financial analysis, the case is a lot stronger.

We go deeper on this exact tradeoff in our Claude Fable 5 vs Opus 4.8 comparison if you want the side by side.

Fable 5 versus Mythos 5

There's a second name floating around that confuses people: Mythos 5. Here's the relationship. Fable 5 and Mythos 5 are the same underlying model. The difference is safety scope.

Fable 5 ships with safety classifiers that hand three kinds of requests back to Opus 4.8 instead of answering them: cybersecurity, biology and chemistry, and distillation (training other models on Fable's outputs). Mythos 5 is the version with those safeguards lifted, and Anthropic restricts it to vetted cyberdefense and biology partners. So if you're using the public model, you're using Fable 5, and you may occasionally get a response that came from Opus 4.8 because a classifier kicked in.

What the safety picture looks like

Since Fable 5 is the most capable Claude that's been made broadly available, the safety details matter more than usual. A few specifics Anthropic published:

• Those handoff classifiers trigger in under 5% of sessions on average. For most everyday work you won't notice them at all.
• External red teaming ran more than 1,000 hours of testing and found no universal jailbreaks.
• Data is kept on a 30 day retention window and is not used for training.

None of that is a guarantee that the model never makes a mistake. No model clears that bar. But the red teaming hours and the 30 day, no-training data policy are the kind of concrete commitments worth knowing before you put sensitive work through any model.

What it costs and where you can get it

Pricing is simple to state:

• $10 per million input tokens.
• $50 per million output tokens.

That's double Opus 4.8, as mentioned. It's also less than half the price of the earlier Claude Mythos Preview, so relative to the preview-tier pricing this is a meaningful drop. We break down the math and what it means for a real monthly bill in our Claude Fable 5 pricing guide.

You can reach Fable 5 in three places:

• The Claude API, using the model id 'claude-fable-5'.
• Consumption based Enterprise.
• AWS Bedrock.

There's also a free window. From June 9 to June 22, 2026, Fable 5 is included in Pro, Max, Team, and seat based Enterprise plans at no extra cost. After June 22, using it on those plans needs usage credits until Anthropic has the capacity to lift that. One honest caveat: that window is time limited and Anthropic could change the dates or terms, so don't build a plan around free access lasting forever.

Is this built for creators?

Worth being clear here, because we write for creators and people will ask. Anthropic did not announce any specific features for writers or creators. The whole release is framed around engineering, science, finance, and enterprise work. The vision and long-context gains could be handy for anyone wrangling big documents or screenshots, but that's a side effect, not the pitch.

So if you're a content creator wondering whether Fable 5 changes your workflow, the honest take is: maybe at the edges, but this launch wasn't aimed at you. It's a model for people shipping large codebases and deep analytical work.

Frequently asked questions

What is Claude Fable 5?

Claude Fable 5 is Anthropic's flagship AI model, released June 9, 2026. It's the first publicly available model in Anthropic's 'Mythos-class' tier, which Anthropic describes as the most capable Claude it has made generally available. It posted state of the art results on most of Anthropic's benchmarks across coding, finance, vision, and long context.

How is Claude Fable 5 different from Opus 4.8?

Anthropic says Fable 5 beats Opus 4.8 on nearly all benchmarks, with the biggest gains on long-horizon complex tasks, finance, and coding. The tradeoff is cost: Fable 5 is priced at double Opus 4.8. Fable 5 also hands a small set of request types (cybersecurity, biology and chemistry, and distillation) back to Opus 4.8 through safety classifiers.

How much does Claude Fable 5 cost?

It's $10 per million input tokens and $50 per million output tokens. That's twice the price of Opus 4.8, and less than half the price of the earlier Claude Mythos Preview.

Is Claude Fable 5 available to the public?

Yes. It's on the Claude API as 'claude-fable-5', on consumption based Enterprise, and on AWS Bedrock. From June 9 to June 22, 2026, it's also included in Pro, Max, Team, and seat based Enterprise plans at no extra cost, after which those plans need usage credits to keep using it. That free window is time limited and could change. If you want a walkthrough of getting started, see how to use Claude Fable 5.