Claude Fable 5

Claude Fable 5 is Anthropic's latest and most capable model, released on 9 June 2026. Anthropic describes it as 'a Mythos-class model that we've made safe for general use', built for the hardest knowledge work and coding problems. It is positioned above Claude Opus 4.8 and is designed to handle long, complex, multi-stage tasks where it plans, delegates to sub-agents, and checks its own work.

No, but they share the same underlying model. According to Anthropic, Fable 5 is the general release with safety classifiers active, while Mythos 5 is a restricted version, available only to approved Glasswing partners and select biology researchers, with some cyber safeguards lifted. In Anthropic's words, 'the safeguards are what distinguish the two models.'

Claude Fable 5 is priced at $10 per million input tokens and $50 per million output tokens, roughly double the cost of Claude Opus 4.8. Prompt caching discounts are available on the API to lower the effective input cost. It is the most expensive Claude model Anthropic has made generally available.

Yes, on Anthropic's own benchmarks. Anthropic says Fable 5 is 'state-of-the-art on nearly all tested benchmarks' and that 'the longer and more complex the task, the larger Fable 5's lead.' Opus 4.8 remains the model Fable 5 falls back to for restricted requests, and it is still a strong, lower-cost choice for everyday work.

Anthropic highlights software engineering, knowledge work, vision, and scientific research, with a particular edge on long-running autonomous tasks. It is built to work across multi-day projects, planning stages, delegating to sub-agents, running and testing its own code, and reading diagrams, charts and PDFs. The harder and longer the task, the more its advantage shows.

Anthropic explains that 'Fable is from the Latin fabula, that which is told, akin to the Greek mythos.' The naming ties Fable to its sibling release, Mythos 5: the two are the same class of model, separated mainly by which safeguards are active. Fable is the public-facing, safety-classified version.

Anthropic released Fable 5 with safety classifiers always on. When the model detects requests in three high-risk areas, offensive cybersecurity, dual-use biology and chemistry, and attempts to distil its capabilities, it automatically routes the answer to Claude Opus 4.8 instead. Anthropic says more than 95% of Fable sessions involve no fallback at all.

It won't answer with the Fable model. Anthropic built in classifiers that detect offensive-cyber, dual-use bio/chem, and capability-distillation requests and silently fall back to Claude Opus 4.8, which has stricter limits in those domains. This routing is the main reason the public 'Fable' version is considered safe for general release, unlike the restricted Mythos 5.

Yes, for a limited time. Anthropic says it requires '30-day retention for all traffic on Mythos-class models' on both first- and third-party surfaces. It states it won't use this data to train new Claude models 'or for any non-safety-related purpose', only to defend against novel attacks and jailbreaks and to reduce false positives.

Claude Fable 5 is available immediately through the Claude API under the model name claude-fable-5, and through consumption-based Enterprise plans. It is also offered via cloud partners such as AWS, Google Cloud and Microsoft Foundry. For subscribers, it is rolling out inside the Claude apps in stages.

Temporarily, for subscribers. Anthropic is including Fable 5 at no extra cost in Pro, Max, Team and seat-based Enterprise plans from 9 June through 22 June 2026. From 23 June, using it requires usage credits. On the API it is paid from day one at $10/$50 per million tokens.

As of mid-2026 the Claude line spans, from fastest to most powerful: Claude Haiku 4.5 (speed and low cost), Claude Sonnet 4.6 (balanced, the workhorse for scale), Claude Opus 4.8 (deep reasoning and coding), and now Claude Fable 5, the new Mythos-class tier sitting on top. Mythos 5 is the same top model with safeguards lifted, restricted to approved partners.

For the hardest, longest coding tasks, Claude Fable 5 now leads, Anthropic built it for large migrations and multi-day sessions where it tests its own work. For most day-to-day development, Claude Opus 4.8 remains excellent and far cheaper, while Sonnet 4.6 is a strong pick for high-volume, cost-sensitive coding. The right choice depends on task length and budget.

Opus is Anthropic's high-capability tier for deep reasoning and complex coding, while Sonnet is the balanced, faster, cheaper tier built for scale. In mid-2026 that means Opus 4.8 for the toughest reasoning and Sonnet 4.6 as the everyday workhorse. Above both now sits Fable 5, and below them Haiku 4.5 handles the fastest, lightest tasks.

Claude Fable 5 is made by Anthropic, the AI safety company behind the entire Claude family. It was released on 9 June 2026 as the public, safety-classified version of Anthropic's Mythos-class model. Anthropic positions itself around AI safety, which is reflected in Fable's built-in fallback to Opus 4.8 for high-risk requests.

Claude Fable 5 was released to the public on 9 June 2026. It is the first generally available version of the Mythos-class model that Anthropic had previously kept to approved partners. The Mythos line itself was first previewed in April 2026 through a restricted program called Project Glasswing.

Project Glasswing is the restricted preview program through which Anthropic first introduced its Mythos-class model in April 2026, before the public Fable 5 release. It involved a small set of approved partners and critical-infrastructure organisations. Mythos 5, the safeguard-lifted version of the model, remains limited to Glasswing partners and select researchers.