Anthropic Launches Claude Fable 5, Its Most Powerful Public Model: What It Means for Operators

June 10, 2026. On June 9, Anthropic launched Claude Fable 5, the most powerful model it has ever made generally available, alongside a restricted sibling called Claude Mythos 5. For the founders, agencies, and small and mid sized teams we work with, the headline is simple: a big jump in long horizon, agentic capability is now available to everyone, and it arrives with a pricing and safety model worth understanding before you wire it into anything that matters.

What Anthropic launched

1. Fable 5 is a step up, not a tweak. Anthropic calls it a Mythos class model, a tier that sits above its Opus class. It is state of the art on nearly all tested benchmarks across software engineering, knowledge work, vision, and scientific research, and Anthropic notes the lead grows as tasks get longer and more complex. In other words, the harder the job, the more it pulls ahead.
2. The proof points are concrete. In early testing, Stripe reported that Fable 5 performed a codebase wide migration in a 50 million line codebase in a single day, work that would have taken a team more than two months by hand. It posts top scores on independent coding and finance benchmarks, and it is the new state of the art on vision tasks, including rebuilding a web app from screenshots alone.
3. Pricing is set. Both Fable 5 and Mythos 5 cost 10 dollars per million input tokens and 50 dollars per million output tokens, which Anthropic says is less than half the price of its earlier Mythos Preview. Developers call it as claude-fable-5 on the Claude API.
4. Availability rolls out in stages. On the API and consumption based Enterprise plans, Fable 5 is fully available now. On subscription plans, it is included free on Pro, Max, Team, and seat based Enterprise from launch through June 22. From June 23, using it on those plans will require usage credits, and Anthropic says it aims to fold Fable 5 back into standard plans once it has the capacity.
5. There are real safeguards. Because a model this capable carries real risk, queries that Anthropic's classifiers flag as related to cybersecurity, biology and chemistry, or model distillation are answered by its next most capable model, Claude Opus 4.8, instead, and the user is told when that happens. Anthropic says more than 95 percent of sessions involve no fallback at all. It also introduced a 30 day data retention policy for business traffic on Mythos class models, which it says will not be used to train models.
6. Mythos 5 is the gated twin. It is the same underlying model with the cybersecurity safeguards lifted, restricted to vetted cyber defenders and infrastructure providers through Project Glasswing, in collaboration with the US government. Most businesses will never touch Mythos 5, and that is the point: Anthropic is splitting the frontier into a public, safeguarded model and a restricted, unrestricted one.

What it means for operators

The practical takeaway is that there is now a clearly best in class option for your hardest, longest tasks. If you have a gnarly code migration, a multi step research or analysis job, or a vision heavy workflow that earlier models fumbled, Fable 5 is the model to reach for. The Stripe result is the kind of thing that turns a quarter long project into a week, and that compression is where the real return on AI lives for a small team.

That said, this is a high end tier, and the smart move is not to run everything on it. At 10 dollars per million input and 50 per million output, Fable 5 is far pricier than mainstream models built for routine work, and the free window on subscription plans closes on June 22, after which it draws on usage credits. Treat it as a precision tool. The winning pattern, the same one Apple effectively endorsed at its developer conference this week by opening the iPhone to multiple AI providers, is model agnostic routing: send the hardest long horizon work to a model like Fable 5, and keep the high volume, routine tasks on cheaper, faster models. Building automations that pick the right model for each job is exactly the kind of AI automation that keeps your costs sane while still giving you frontier capability where it counts.

If you build on the API directly, design around two details. First, the safeguard that reroutes some security, biology, and chemistry queries to Opus 4.8: it fires in under 5 percent of sessions and tells the user when it does, so most business workflows will never notice, but if your product touches those topics, plan for it. Second, the new 30 day data retention requirement for Mythos class traffic, which matters for compliance and data governance even though Anthropic says the data is not used for training. These are not blockers, they are simply the fine print of running on the frontier, and getting them right is part of why teams bring in an experienced engineer rather than wiring it up blind. If you want that senior help, you can hire an AI engineer from us to integrate Fable 5 cleanly into your product or operations.

The bigger picture is the one worth sitting with. Capability is compounding faster than most businesses can absorb it, and the gap is no longer between companies that have AI and companies that do not. It is between teams that have operationalized these models inside real workflows and teams still running one off experiments. Fable 5 widens that gap. For founders building a product, the right response is to treat frontier models as core infrastructure from day one, which is the approach we take when we help teams launch a SaaS. The launch is loud, but the advantage goes quietly to whoever puts the capability to work first.