Anthropic has launched Fable 5 and Mythos 5, two new Claude models, with distinct capabilities and pricing aimed at enhancing AI applications.
On June 9, Anthropic introduced two new versions of its Claude model family: Fable 5 and Claude Mythos 5. Fable 5 is designed for general use, while Mythos 5 operates as a more restricted, private model.
The primary distinction between the two lies in their protective classifiers. The Fable model actively monitors sensitive topics including cybersecurity and biology, rerouting certain queries to Claude Opus 4.8. This safety measure, while conservative, has been noted to trigger on less than 5% of sessions, according to the company.
"To release the model safely and quickly, we configured these protective measures conservatively — sometimes they respond to harmless queries," Anthropic stated.
Fable 5 is accessible via a subscription through the web interface and Claude API, with a pricing structure of $10 per million input tokens and $50 per million output tokens, which is significantly lower than the previous Claude Mythos Preview model.
Details on Mythos 5 Access
Mythos 5 has been made available to a select group of trusted participants, including those involved in Project Glasswing, and replaces the earlier Claude Mythos Preview. Anthropic plans to broaden access for specific tasks in cybersecurity and biomedical research by late 2026.
While both models come with restrictions, Mythos 5 is touted as the most capable for cybersecurity, demonstrating proficiency in identifying software vulnerabilities and expediting certain drug development processes.
"Mythos 5 utilizes protein design and bioinformatics tools, which often outperform experienced human operators," the company noted, highlighting promising outcomes in drug development research.
For biomedical researchers, a distinct access route to Fable 5 will be established, allowing more freedom in biological and chemical inquiries while still maintaining cybersecurity restrictions.
Performance Highlights of Fable 5
Anthropic claims that Fable 5 is the strongest model available to the public, excelling in complex tasks such as programming, document analysis, and scientific research. However, the company has not disclosed specific details about the model’s architecture or parameters.
Preliminary tests by Stripe indicated that Fable 5 significantly reduced development time, completing a complex migration of a 50 million line Ruby codebase in just one day, a task that would typically require two months.
Fable 5 has also demonstrated enhanced token efficiency compared to earlier Claude versions, showing superior performance in various evaluations, including the FrontierCode and Hebbia Finance Benchmark tests.
One notable feature of Fable 5 is its advanced "vision" capability, allowing it to extract data from intricate scientific diagrams and execute complex tasks with minimal assistance. The model can effectively maintain focus over long tasks, enhancing its performance by utilizing internal notes.
Recently, Anthropic cautioned about the potential risks associated with AI self-improvement, revealing that over 80% of the code for its current products was generated by Claude itself.
