Anthropic Opens Its Most Advanced AI to Everyone—With Strings Attached

Anthropic is giving the general public its first taste of Mythos, the company's most powerful AI model, but it's not letting it run wild. On Tuesday, the firm released Claude Fable 5, a version of Mythos that anyone can use through the Claude API or consumption-based Enterprise plans. Fable 5 is built for heavy lifting—software engineering, knowledge work, and vision tasks—but it comes with hard safety limits. In sensitive areas like cybersecurity, biology, chemistry, and model distillation, the system refuses to answer and hands off to the less capable Claude Opus 4.8.

Mythos first appeared as a preview in April, locked down to a handful of partners because of security worries. Last week, Anthropic expanded access to hundreds of organizations in 15 countries, mostly those running critical infrastructure. Now, a dialed-back version is open to the public. For a limited time—through June 22—Fable 5 is included in Pro, Max, Team, and seat-based Enterprise plans at no extra cost. After that, users will need to pay with usage credits until Anthropic adds it back as a standard feature.

Anthropic is also rolling out Mythos 5 to organizations that already had approval for the full model. The launch comes as the company eyes public markets, alongside OpenAI and SpaceX. It also follows Anthropic's call for major AI labs to agree on an emergency brake for frontier systems. The company warns that models could soon reach recursive self-improvement, where they upgrade themselves without human help.

Before releasing Fable 5, Anthropic tested its defenses hard. An external bug bounty ran over 1,000 hours without finding a single universal jailbreak. External red teams also failed. Still, new attacks could emerge. So Anthropic now requires 30-day data retention on all traffic—even for enterprises that previously had zero-retention deals. The company says it won't use the data for training, only to catch novel attacks and reduce false alarms. The policy could set a new industry norm: more powerful models, more surveillance.

For most users, Fable 5 handles the job. Early data shows at least 95% of sessions run entirely on the model. In third-party tests, analytics firm Hex gave Fable a 90% on complex, long-running analytical tasks. Vibe-coding platform Base44 praised its ability to build full apps in one go. Genspark said Fable beat every other model in its evaluations, especially on UI design and game coding.

Pricing is steep: $10 per million input tokens and $50 per million output tokens—double the cost of Opus 4.8. That alone may keep casual users away. But companies like Rakuten see the value. “At the highest effort, Fable reflects on and validates its own work,” Rakuten said. “For us, that’s what makes highly autonomous operations possible—the extra thinking pays for itself.”