Anthropic just made its most powerful AI model available to the public. The new version is called Claude Fable 5. It belongs to the Mythos class, a group of highly advanced AI systems.

Previously, Mythos models were tightly controlled and available only to a few trusted partners. Now, anyone with a subscription or pay-per-use plan can access Fable 5. Anthropic designed this version with safety in mind.

That means Fable 5 has built-in guardrails. For sensitive topics like cybersecurity, biology, and chemistry, the model routes those queries to a safer, less capable AI called Claude Opus 4.8. This helps prevent misuse without cutting off access to the full power of Mythos.

Fable 5 delivers top-tier performance across many tasks. It excels in coding, reasoning, knowledge work, and even vision-based challenges like interpreting scientific images. This model can handle long and complex work sessions better than its predecessors.

Performance and Safety Combined

During tests, Fable 5 outperformed older Claude models and even some competitors. Financial and analytical benchmarks showed major improvements. For example, it can analyze documents, interpret charts, and solve tough problems with strong accuracy.

In software engineering, the model made waves by migrating a massive 50-million-line codebase in just one day. That’s a task that typically takes months for a team of engineers. It also showed great skill in creating user interfaces and coding games.

Despite its power, Anthropic took safety seriously. The company ran extensive tests before release, including over 1,000 hours of bug bounty hunting. External security teams found no universal jailbreaks to bypass the safeguards.

Still, Anthropic is cautious. It enforces a 30-day data retention policy for all traffic, even for enterprise clients who previously had zero-retention agreements. This data won’t train future models but helps defend against new attacks and improve safety filters.

Access, Pricing, and Future Plans

Fable 5 is currently included in all Claude subscription tiers until June 22. After that, users will need to buy usage credits to continue accessing the model. The pricing is $10 per million input tokens and $50 per million output tokens.

That’s about twice the price of the safer Claude Opus 4.8 model. The higher cost may slow adoption, but some companies like Rakuten say the model’s autonomy and accuracy justify the expense.

Meanwhile, Anthropic is also releasing Mythos 5, a less restricted version of the model. Mythos 5 is available only to approved cybersecurity organizations and infrastructure providers. This version offers fewer limitations but remains tightly controlled due to its potential risks.

Anthropic plans to expand Mythos 5 access through a trusted partner network in the future. The company also supports calls for the AI industry to agree on slowing down frontier AI development when needed. They warn that AI could soon improve itself without human help, making control harder.

Anthropic’s move to open Mythos-class AI marks a shift. Powerful AI is no longer confined to a select few. But it comes with strict safety measures and new pricing models. This balance between access and caution could set a new standard for AI deployment.

As AI continues to evolve, models like Claude Fable 5 will shape how businesses and developers use advanced tools. They offer incredible capabilities but also demand careful oversight to avoid misuse.