[DigitalToday reporter Chi-gyu Hwang (황치규)] Anthropic has launched a public version of its Mythos model, branded as 'Claude Fable 5,' TechCrunch reported on June 9 (local time).

According to the report, Fable 5 shows strong performance in software engineering, knowledge work and vision. In high-risk areas such as cybersecurity, biology, chemistry and distillation, it blocks responses and switches to Claude Opus 4.8. Mythos, released as a preview in April, was initially limited to a small number of partners over concerns about its impact on cybersecurity, before being expanded last week to 150 organizations in 15 countries.

The announcement makes it available to anyone through Anthropic's Claude API and a usage-based enterprise plan.

Until June 22, existing Pro, Max, Team and Enterprise subscribers can use Fable 5 at no additional cost. From June 23, it will be excluded from subscription plans and users will need to buy credits separately. API pricing is $10 per 1 million input tokens and $50 per 1 million output tokens, about double the level of Opus 4.8.

Anthropic said it will be required to retain all Fable 5 traffic for 30 days, including for companies that previously had agreements not to store conversation content. It said the goal is to analyse new jailbreak attacks and reduce false positives.

Anthropic said it conducted more than 1,000 hours of bug bounty testing before launch, and that external red-team groups also could not find a way to jailbreak the model. Early data showed more than 95 percent of Fable sessions were handled solely by Fable responses without switching to Opus.

Analytics company Hex said Fable was the first to reach 90 percent on its core analytics benchmark, and AI workspace platform Genspark said Fable far outperformed all other models in UI design and game coding.