Anthropic’s Claude Fable 5 is the Mythos version that the public can access today

Anthropic is bringing its most powerful AI model to the general public for the first time, but it’s doing so with caution.

On Tuesday, the AI ​​company launched Claude Fable 5, the first publicly available version of its Mythos model. Anthropic says the Fable 5 excels in software engineering, information processing, and vision, but comes with strict security limitations. In high-risk areas such as cybersecurity, biology, chemistry, and distillation, the model blocks responses and reverts to Claude Opus 4.8.

Launched as a preview in April, Mythos was initially limited to a few partners due to cybersecurity concerns. Last week, Anthropic expanded its reach to hundreds of organizations in 15 countries, with a focus on organizations that manage critical infrastructure.

Now, a version of that technology is available to anyone through Anthropic’s Claude API and application-based Enterprise plans. Subscription access will be rolled out in stages: until June 22, Fable 5 is included in the Pro, Max, Team, and Enterprise plans based on the chair at no additional cost. On June 23, Anthropic will release Fable 5 on those plans, which require credits to use going forward, with plans to bring it back as a standard subscription feature as soon as possible.

Anthropic is also shipping a new version of Mythos, called Mythos 5, to organizations that are already authorized to access the upgraded model.

The launch of Fable comes as Anthropic prepares to enter the public markets, alongside OpenAI and Elon Musk’s SpaceX. It also follows an AI company’s request urging major AI labs around the world to establish a brake linked to cross-border AI development. Anthropic warned that systems are evolving so rapidly that they may achieve iterative self-improvement (RSI), improving themselves independently without human intervention.

Aware of how a Mythos-class model can do in the wrong hands, Anthropic says it tested stress classifiers with jailbreak attempts before releasing Fable 5.

“Internally, we had an external exploit for a bug that did not produce a global jailbreak in over 1,000 hours of testing. We then worked with external red-teaming organizations that also failed to find a global jailbreak.”

That said, there could still be novel attacks that could happen. Because of this, with the launch of Fable 5 and Mythos 5, Anthropic said it will require 30-day retention on all traffic, even if businesses previously had no-retention agreements. Anthropic said it would not use the data for training, only “to protect against sophisticated and novel attacks, including new jailbreaks,” and to “identify and mitigate false positives.” The policy could set an industry precedent where access to increasingly powerful models comes with mandatory data retention policies installed as a security measure.

For those who continue to use the model, not every question will receive a Fable 5 answer. Anthropic says that the situations where Fable has to defer to Opus 4.8 are rare, with early data showing at least 95% of Fable sessions working completely on the model’s own answers.

In third-party testing, analytics firm Hex said in a statement that Fable was the first to score 90% on its baseline test for complex, long-running analytics tasks.

“On the most difficult questions, it shows strong judgment and little attention,” Hex said.

Vibe-coding platform Base44 noted in a statement that Fable is better for “full-featured one-shot applications” and has better tooling. AI-powered workstation and agent platform Genspark said Fable outperformed all other models in its analysis, and performed significantly better in tasks such as UI design and game coding.

The price of both Fable 5 and Mythos 5 is $10 per million input tokens and $50 per million output tokens, double the price of Opus 4.8. That price alone may act as a deterrent to widespread use.

Many businesses are increasingly critical of AI costs after seeing bills come in or out of their annual AI budget early on. Advanced models like Opus 4.8 can increase those problems, with advanced processing capabilities that can split a single application into multiple tasks.

Anthropic said it expects demand for Fable 5 to be very high and difficult to predict. And others, like shopping rewards platform Rakuten, may think the upside is worth the price.

“With the greatest effort, Fable demonstrates and validates its work,” Rakuten said in a statement. “For us, that’s what makes the most independent jobs possible – extra thinking pays for itself.”