Anthropic released Claude Fable 5 on 9 June — a Mythos-class model now open to all developers and enterprises, state-of-the-art on complex autonomous tasks, and priced at $10/$50 per million tokens, less than half the cost of Claude Mythos Preview.

Fable 5 is the model Anthropic has been building toward. Released on 9 June, it is the first time the company has made a Mythos-class model — the tier above its flagship Opus range — generally available. The release marks a significant step in what AI tools can do for knowledge workers: Fable 5 is built for the long, complex, autonomous tasks that previous models struggled to complete reliably, and it arrives at a price point that makes it accessible to the broad developer and enterprise market rather than a restricted circle of partners.

The new model is available via the Claude API at $10 per million input tokens and $50 per million output tokens. That is less than half the price of Claude Mythos Preview, the Mythos-class model released in April through Project Glasswing, which had been restricted to a small group of cybersecurity partners and infrastructure providers. Alongside Fable 5, Anthropic released Claude Mythos 5 — the same underlying model with some safety classifiers lifted — for the same Glasswing partners, now at the same reduced pricing.

What Fable 5 can do that previous models could not

The capabilities that set Fable 5 apart from Anthropic’s Opus-class models are concentrated in long-horizon, autonomous work: tasks that require sustained focus, complex reasoning and the ability to recover from failures over extended sessions, rather than a single sharp response.

Stripe, which had early access, reported that Fable 5 compressed months of engineering work into days. In a 50-million-line Ruby codebase, the model completed a codebase-wide migration in a day that would otherwise have taken a whole team more than two months by hand. On Cognition’s FrontierCode evaluation — which tests whether models can pass difficult coding tasks while meeting production codebase standards — Fable 5 scores highest among frontier models.

Knowledge work performance is equally striking. On Hebbia’s Finance Benchmark for senior-level reasoning, Fable 5 achieved the highest score of any model tested, with substantial gains in document-based reasoning, chart and table interpretation and complex problem-solving. IMC, the trading firm, noted that the model aced its trading-analysis evaluations nearly across the board, including factual lookup, conceptual reasoning, root-cause analysis and expected-value analysis.

Vision and memory capabilities have also advanced materially. Fable 5 can extract precise numbers from detailed scientific figures and rebuild a web application’s source code from screenshots alone. In long-running sessions, it stays focused across millions of tokens and uses its own notes to improve its outputs as a task progresses.

The safeguard mechanism and what it means in practice

Releasing a Mythos-class model for general use required Anthropic to solve a problem it had flagged in April: models at this capability level carry meaningful dual-use risk, particularly in cybersecurity and biology. The solution is a classifier system that detects queries related to those areas and routes them to Claude Opus 4.8 instead.

When this happens — on classified queries covering cybersecurity, biology and chemistry, or distillation attempts — users are notified and receive an Opus 4.8 response. Anthropic’s early data shows the fallback triggers in fewer than 5 per cent of sessions. For the remaining 95 per cent or more, Fable 5’s performance is effectively equivalent to Mythos 5’s.

The practical implication for most B2B users is that the classifier will rarely be visible. Marketing teams, knowledge workers, developers building business applications and analysts using the model for research and strategy work are unlikely to encounter the boundary regularly. The classifiers have been tuned conservatively by design — Anthropic has acknowledged this means some legitimate requests in adjacent domains will trigger a fallback, and has committed to reducing false positives as the safeguards are refined.

Pricing, availability and what changes for builders

Fable 5 is available immediately via the Claude API using the model string claude-fable-5. For subscription plans — Pro, Max, Team and seat-based Enterprise — Anthropic is rolling out access in stages, with the model included at no extra cost from 9 June through 22 June, after which usage credits will be required until capacity allows it to be restored as a standard plan feature.

The pricing is the clearest signal of intent. Mythos Preview was priced for a restricted partner cohort; Fable 5 at $10/$50 per million tokens is priced to be deployed at scale. For developers building AI-assisted workflows, agents and knowledge tools on the Claude API, this is the model that makes Mythos-class capabilities economically viable for production applications — the long autonomous tasks, complex document reasoning and multi-step problem-solving that were previously too expensive to run routinely.

Anthropic’s framing positions Fable 5 as the default choice for the hardest tasks: the longer and more complex the task, the larger its lead over previous Claude models. For B2B marketing and knowledge teams, the relevant implication is that the autonomous AI work that has so far remained experimental — agent-driven research synthesis, complex content workflows, multi-document analysis at scale — moves closer to practical deployment with Fable 5 as the underlying model.