5 things to know about Claude Fable 5

Anthropic has released an AI model it says is more capable than anything it has previously made generally available, while also admitting that releasing it carries risks. Claude Fable 5 is the public-facing version of the same underlying model behind Claude Mythos 5, with extra safeguards placed around cybersecurity, biology, chemistry and distillation. For Australian readers, the launch matters because Claude Mythos access has already reached Australia through Project Glasswing, bringing both cyber-defence promise and a sharper divide between organisations that can use advanced AI safely and those that cannot.

What We Know So Far

Anthropic says Claude Fable 5 is available everywhere today, while Claude Mythos 5 remains restricted to approved partners and selected trusted-access users. Both are described as the same underlying model, but Fable 5 has safeguards that redirect some sensitive requests to Claude Opus 4.8 instead of allowing the more capable model to answer directly.

The company says those safeguards apply to areas including cybersecurity, biology and chemistry, and distillation. Anthropic says the fallback system triggers in less than 5% of sessions on average, meaning more than 95% of Fable sessions run without fallback. The practical trade-off is clear: the public gets access to a much stronger model, but some harmless requests may be caught by conservative filters.

Pricing is another signal that this is not a casual upgrade. Fable 5 and Mythos 5 are priced at $10 per million input tokens and $50 per million output tokens. Anthropic says Fable 5 is included on Pro, Max, Team and seat-based Enterprise plans at no extra cost from launch through June 22. From June 23, using it on those plans will require usage credits unless capacity allows an extension.

The model is also appearing quickly through major developer and enterprise channels. GitHub says Claude Fable 5 is available in GitHub Copilot for Pro+, Max, Business and Enterprise users, with administrators on Business and Enterprise plans required to enable a policy that is off by default. Microsoft says Fable 5 is available in Microsoft Foundry, where it is being positioned for autonomous agents, complex coding, document-heavy workflows and enterprise governance.

Voices & Opinions

Anthropic’s own framing is unusually direct for a product launch. It presents Fable 5 as a major capability jump, but also as a release that required heavier controls because the same strengths that help defenders and researchers can also help malicious actors.

Fable's capabilities exceed those of any model we've ever made generally available

That tension runs through the wider response. The Conversation’s Australia-focused coverage notes that Mythos has been rolled out to an additional 150 organisations across 15 countries, including the Australian government and several local businesses, as part of Project Glasswing. It also warns that AI-generated vulnerability reports still need human validation, because unprepared security teams can be overwhelmed by false reports.

Enterprise partners are treating access carefully. GitHub says Anthropic retains prompts and outputs for up to 30 days for Fable 5 safety classifiers, while other Claude models in GitHub Copilot continue under zero data retention. Harvey, the legal AI platform, says Fable 5 is opt-in for its customers because its data processing practices differ from existing commitments and because data may be processed in the United States.

Local Impact

For Australia, the immediate impact is strongest in cybersecurity and enterprise AI adoption. The Conversation reports that the Australian government and local businesses are among the Project Glasswing users getting access to Claude Mythos, a tool designed to identify serious bugs and vulnerabilities before attackers can exploit them.

The risk is uneven access. Large organisations with specialist teams may be able to use Mythos or Fable safely, triage findings and patch systems quickly. Smaller Australian businesses may not have the same staff, budget or governance processes. If AI speeds up vulnerability discovery, then slow patching becomes more dangerous because attackers may also gain stronger tools over time.

For developers in Australia, Fable 5’s arrival in GitHub Copilot and Microsoft Foundry means the model may appear inside familiar coding and enterprise platforms rather than only through Anthropic’s own API. That convenience comes with governance decisions: business administrators must consider data retention, cost, model access policies and whether sensitive work should be routed through a model that requires safety monitoring.

Coming Up

Anthropic says Fable 5 is available now through the Claude API and consumption-based Enterprise plans, with subscription access included only through June 22 unless capacity allows a longer included period. It also plans to expand Mythos 5 through a broader trusted access program, including a biology access program for selected researchers where biology and chemistry safeguards are removed while cyber safeguards remain in place.

Developers and business users should also watch platform-specific rollouts. GitHub says rollout in Copilot will be gradual, and Copilot Business and Enterprise administrators must enable the Fable 5 policy before users can select it.

At a Glance

• Claude Fable 5 is Anthropic’s public version of a Mythos-class model.
• Claude Mythos 5 uses the same underlying model but has fewer safeguards for approved trusted users.
• Fable 5 redirects some cybersecurity, biology, chemistry and distillation requests to Claude Opus 4.8.
• Anthropic says fallback happens in less than 5% of sessions on average.
• Pricing is $10 per million input tokens and $50 per million output tokens.
• Australian government and local businesses are among Project Glasswing users reported to have access to Mythos.