The new artificial intelligence model **Claude Fable 5** has been launched as a Mythos-class model configured for general public use.
This release represents a significant shift in available computational capabilities, outperforming previous models on multiple technical benchmarks. Specifically, the developer stated that the model is offered alongside Claude Mythos 5, which is a version of the same underlying model designed for specialized cyberdefenders with certain safeguards removed.
Technical Capabilities of Claude Fable 5
In early testing, Stripe reported that **Claude Fable 5** compressed months of engineering work into days by performing a codebase-wide migration on a 50-million-line Ruby codebase in a single day. Furthermore, the model achieved the highest score among frontier models on Cognition’s FrontierCode evaluation, which measures production-grade coding capabilities.
In addition to software development, the model excels at complex analytical tasks. Notably, it scored highest on Hebbia’s Finance Benchmark for senior-level reasoning, showing significant gains in document-based reasoning and chart interpretation. Meanwhile, the model also established a new benchmark in vision tasks, demonstrating the ability to rebuild web application source code directly from screenshots.
Specialized Scientific Research and Drug Design
The underlying technology is also being utilized in health technology through Claude Mythos 5. In internal testing, protein design experts used the model to accelerate aspects of the drug design process by approximately ten times. Specifically, the model successfully executed tasks normally completed by scientists, such as choosing binding sites and running protein design tools without human assistance.
Moreover, the model conducted autonomous genomics research over a week-long period. It assembled single-cell data for millions of cells across 138 animal species and trained a custom machine learning model. Consequently, this trained model outperformed a recently published model in the journal Science, despite being 100 times smaller in size.
Safety Classifiers and Fallback Systems
To manage the risks associated with these advanced capabilities, the developer integrated new safety classifiers. When the classifiers detect queries related to cybersecurity, biology, chemistry, or model distillation, the system automatically routes the request to Claude Opus 4.8. According to early data, these safeguards trigger in less than 5% of user sessions on average.
The cybersecurity classifiers are designed to block offensive cyber tasks and exploitation attempts. In testing, the model complied with zero harmful single-turn requests regarding cyberattack planning or exploit development. Additionally, a new data retention policy requires a 30-day retention period for all traffic on these models to help identify and prevent novel attacks.
Pricing and General Availability
Both models are priced at $10 per million input tokens and $50 per million output tokens, which is less than half the price of the previous preview version. While **Claude Fable 5** is available immediately via the API and Enterprise plans, subscription users on Pro, Max, and Team plans will have access included at no extra cost until June 22, 2026.
After June 23, 2026, subscription users will need to use usage credits to access the model. Meanwhile, access to Claude Mythos 5 remains restricted to Project Glasswing partners and select biology researchers through a trusted access program. The developer plans to expand these programs to more organizations in the future.
