Is Claude Fable 5 Worth It? Pricing After 22 June, vs Opus 4.8 and ChatGPT

You have until 22 June to use Anthropic's most capable model for nothing. Then the meter starts.

On 9 June 2026, Anthropic released Claude Fable 5: a new tier above Opus, available in Claude Code from day one, and included free on Pro, Max, and Team plans for exactly thirteen days. From 23 June it moves to prepaid usage credits at double the per-token price of Opus 4.8. That deadline turns a product launch into a decision: is this model worth paying a premium for, or is the one you already have good enough?

We've spent the launch window using it for real work, including building a complete public tool with a single brief. This is what Fable 5 is, what it costs, how it compares to Opus 4.8 and ChatGPT's GPT-5.5, and how to decide. If you're after a consumer recommendation rather than a developer one, our AI subscription comparison for everyday use covers that instead.

What Claude Fable 5 actually is

In April we wrote about Claude Mythos, the model Anthropic built and then refused to release because it could find and chain zero-day exploits on its own. We predicted the reasoning gains would "filter into the product line" without the dangerous tooling. That's exactly what Fable 5 is.

Anthropic's announcement describes two models built from the same base. Claude Mythos 5 keeps the offensive cybersecurity capability and goes only to vetted partners through Project Glasswing. Claude Fable 5 is the public release: identical intelligence, with classifiers that detect restricted topics (offensive security, biology, chemistry, attempts to distil the model) and route those requests to Opus 4.8 instead. Anthropic says over 95% of sessions never touch the classifiers, and calls the fallback "a far better experience than an outright refusal".

The company's own framing is blunt: Fable 5's capabilities "exceed those of any model we've ever made generally available". External coverage from CNBC and The Decoder backs the headline numbers. It has a one-million-token context window, ships in Claude Code, the Claude apps, and the API as claude-fable-5, and it is not a speed upgrade. It's deliberately the opposite.

What it's like to use

On 10 June, the day after launch, we used Fable 5 in Claude Code to build a new free tool for our sister site: the AI Bot Checker. The brief was outcome-shaped rather than a spec: an easy-to-understand checker that tests whether AI bots are blocked from a website, including robots.txt analysis. What it should do, and what using it should feel like. Nothing about how to build it.

Fable read through the existing codebase and the tools already on the site, then built the whole thing. Thirteen AI crawlers tested two ways (robots.txt rules plus live requests sent with each bot's user agent), a bot-by-bot results table, an A to F grade, and a ready-to-send message for your hosting provider when something is blocked. Functionality and design landed right first time. No correction pass. In three years of building with these models, that's not happened to us on a job this size before.

Two things stood out in day-to-day use. It's slower than Opus 4.8 per response, sometimes noticeably. And it reads far more before it writes: more files opened, more existing patterns studied, before the first line of code appears. The slowness buys something. Fewer wrong turns, fewer "no, not like that" corrections, less of your time spent steering.

We're not the only ones seeing this pattern.

"The challenge is finding tasks that it can't do" is the line that made me put my coffee down, because it matched the day I'd just had. I went looking for the catch in our one-shot build for a good hour: an edge case missed, a layout that breaks on mobile, anything. I didn't find one. Willison also buried a number in his post that matters more than any benchmark: he burned through $110.42 of Fable usage in a single day, all inside his $100-a-month subscription. That's the free window doing its job, and a preview of what the meter looks like when it switches on.

What it costs after 22 June

The pricing has two phases, and the boundary between them is 11:59pm on 22 June 2026.

Right now, Fable 5 is included at no extra cost on Pro, Max, Team, and seat-based Enterprise plans. It draws from your normal usage limits, and because it's a bigger model, it draws faster. From 23 June it leaves those plans entirely: continued use requires prepaid usage credits billed at API rates. Anthropic says it intends to bring Fable back into standard plans "once capacity grows", with no date attached.

The sticker shock needs one qualifier. Per-token price is not per-task price. A model that solves a job in one attempt can cost less overall than a cheaper model that needs three runs and an hour of your corrections. Finout's cost analysis makes the same point from the other direction: for routine, well-defined work like summarising and extraction, the 2x premium over Opus 4.8 is hard to justify. Both things are true. The premium pays for itself on hard jobs and wastes money on easy ones.

For scale: we run the Max 20x plan at roughly £180 a month, and Fable's free window has been absorbing work that would have been billed at well over that on the API. Willison's $110.42 day points the same way. If your usage looks anything like that, the post-22-June question isn't whether to buy credits. It's which jobs deserve them.

Fable 5 vs Opus 4.8 vs GPT-5.5

OpenAI's current flagship is GPT-5.5, released in April 2026 at $5/$30 per million tokens. Here's how the three stack up on the published numbers.

The FrontierCode row is the one to stare at. SWE-bench Pro scores cluster within 22 points of each other, but on the hardest coding evaluation in current use, Fable 5 more than doubles Opus 4.8 and posts five times GPT-5.5's score. That's the shape of a model built for long, difficult jobs rather than quick ones. GPT-5.5 still wins on response speed and runs cheaper, which is why plenty of developers will reasonably keep it for fast iterative work. We compared the previous generation in our sister site's Claude vs GPT vs Gemini guide, and the pattern holds: each vendor's strengths sit on a different axis.

Not everyone reads Anthropic's tiering charitably, and the loudest sceptic runs the competition.

He said that in April, about the Mythos rollout that Fable 5 now extends, and I'll admit it lodged in my head while writing this. The cynical reading writes itself: invent a scary tier, gate it, then sell the safe version at double price. What softened my cynicism was the thirteen free days. A company purely milking fear doesn't hand its most expensive model to every subscriber for a fortnight and let people like Willison burn $110 a day on a $100 plan. That looks less like a shelter sales pitch and more like a company actually short of compute, sampling demand before it prices the queue. Both readings can be partly true at once. Hold them both.

Should you use it?

Until 22 June, the answer is simple: yes. It costs you nothing extra, and the free window is the cheapest evaluation you'll ever run on a frontier model. Point it at your hardest backlog item and see what comes back.

From 23 June, it's a budgeting question, and the honest answer splits by workload:

• Use Fable 5 (on credits) when the job is large, self-contained, and expensive to get wrong: a new tool built end to end, a gnarly migration, a refactor that touches forty files. One-shot accuracy is where the 2x premium earns its keep.
• Stay on Opus 4.8 when the work is iterative, conversational, or routine. It's still an excellent model, it's half the price, and it stays inside your existing plan limits.
• Skip both and look at GPT-5.5 when raw response speed matters more than depth, or your team is already inside the Codex toolchain.

Our own plan after the window closes: Opus 4.8 stays the daily default in Claude Code, and we'll buy credits for Fable when a job looks like the AI Bot Checker did. A big, well-briefed, self-contained build where getting it right first time is worth real money. On current evidence that's perhaps two or three jobs a month, not twenty.

One caveat worth knowing before you commit work to it: the guardrail routing is silent. If your prompt brushes a restricted topic, you get Opus 4.8's answer without being told mid-session. For most development work you'll never hit it. If you work in security, you will, and you should plan around it.

What this means for your website

There's a thread connecting a frontier coding model to the work this site usually covers. The tool Fable built for us in one shot exists because AI crawler access has become something ordinary site owners need to check. Models this capable are increasingly the things reading your website: fetching pages, comparing claims, deciding whether your business is the one to recommend. We walked through that pipeline in how AI search actually works, and our server logs show AI crawlers reading machine-readable files daily.

And the blocking problem is real, not theoretical. The first day we ran the new tool against live sites, one scored a D: ClaudeBot, Bytespider, and CCBot all rejected at server level with 403s, invisibly to the owner. Fable's build even drafts the email to send your hosting company about it.

A model that can one-shot a working web tool can certainly parse your homepage. What it can't do is invent facts you never published. AI Visibility Checking exists to verify that the inputs are there: that crawlers can reach you, and that files like llms.txt and developer-ai.txt state who you are in a form these models can use. The quick-start guide covers the full set in an afternoon. As the models get sharper, ambiguity stops being neutral. It becomes the reason you're left out of the answer.

Frequently asked questions