Anthropic releases Claude Fable 5 to the public with safety limits

In short: Anthropic has released Claude Fable 5, a more powerful version of its Claude AI that the public can use, but with built-in blocks for some risky topics.

What happened

Anthropic launched Claude Fable 5, which it says is the first version of its “Mythos-class” model that is available to the general public. A model is the engine behind an AI chatbot, like the “brain” that produces answers.

People can access Fable 5 through Anthropic’s Claude API (a way for apps and companies to plug into Claude) and through usage-based Enterprise plans. Anthropic said Fable 5 is strong at software work, office style research, and understanding images.

Anthropic also added strict safety limits. For certain topics, including cybersecurity, biology, and chemistry, the system will block the response and instead use an older model, Claude Opus 4.8. Think of it like a car that can go faster, but it automatically slows down on roads where mistakes could cause more harm.

Access is also changing on a schedule. Through June 22, Fable 5 is included at no extra cost in Pro, Max, Team, and seat-based Enterprise plans. Starting June 23, Anthropic plans to require usage credits for Fable 5, and it says it will try to add it back to subscriptions as a standard feature later.

Anthropic set the price at $10 per million input tokens and $50 per million output tokens. Tokens are small chunks of text (roughly like word pieces) that companies pay for as they send prompts in and get answers back.

Why it matters

More powerful AI tools are reaching more people, but companies are also adding stronger controls and even keeping more logs of how the tools are used. Anthropic said it will keep 30 days of traffic data for Fable 5 and Mythos 5 to help spot new “jailbreaks” (tricks that try to bypass safety rules), even for some customers who previously had no data retention.

Source: TechCrunch AI