Anthropic Unleashes Claude Fable 5 and Restricted Mythos 5, Redefining AI Capabilities

Anthropic's Mythos-Class Arrives: Fable 5 for the Public, Mythos 5 for the Elite

Anthropic has officially rolled out its latest generation of AI models, introducing both Claude Fable 5 for general availability and the highly restricted Claude Mythos 5. Both models belong to a new "Mythos-class" of AI, a tier of capability surpassing Anthropic's previous Opus class models. While Fable 5 is designed for broad public use, Mythos 5, which shares the same underlying capabilities without certain safety classifiers, is currently limited to approved customers within Project Glasswing, including the US government and select biology researchers. This dual launch highlights Anthropic's strategy of balancing cutting-edge AI innovation with rigorous safety protocols. The release of Fable 5 follows a period where the raw Mythos model was deemed too powerful for public release due to cybersecurity concerns, with access initially restricted to a handful of trusted organizations. Anthropic has now implemented robust safeguards in Fable 5, which reroute sensitive queries related to cybersecurity, biology, chemistry, and health to the less capable Opus 4.8 model. These classifiers, separate AI systems designed to detect and prevent misuse, trigger in less than 5% of sessions on average.

Unprecedented Performance Across Diverse Domains

Claude Fable 5 is being hailed as state-of-the-art across nearly all tested benchmarks of AI capability, demonstrating exceptional performance in areas such as software engineering, knowledge work, vision, and scientific research. Its lead over previous models, including Opus 4.8, becomes more pronounced with longer and more complex tasks. For instance, Fable 5 scored 91 out of 100 on Every's Senior Engineer benchmark, significantly outperforming Opus 4.8 (63) and GPT-5.5 (62). One of the most engaging demonstrations of Fable 5's capabilities is its ability to generate playable 3D video games from a single sentence prompt. Examples include a "Library of Babel explorer" and a "self-aware version of Snake," showcasing its capacity for long-horizon coding and complex, multi-step software engineering tasks. Beyond game creation, Fable 5 also exhibits advanced vision capabilities, understanding diagrams, charts, and tables within PDFs, and can even rebuild a web app's source code from screenshots alone. It has also shown the ability to autonomously play games like Pokémon FireRed using only vision.

Pricing, Availability, and the Future of Mythos

Claude Fable 5 is now generally available through the Claude API, Claude Platform on AWS, Amazon Bedrock, Vertex AI, and Microsoft Foundry. The pricing for both Fable 5 and Mythos 5 is set at $10 per million input tokens and $50 per million output tokens, which is less than half the price of the earlier Claude Mythos Preview, yet approximately double the cost of Claude Opus 4.8. Anthropic is offering a temporary free window for Fable 5 on Pro, Max, Team, and Enterprise plans until June 22, after which usage credits will be required. The more powerful, unrestricted Claude Mythos 5 is being deployed through Project Glasswing in collaboration with the US government, serving as an upgrade to Claude Mythos Preview. Mythos 5 boasts the strongest cybersecurity capabilities of any model globally and has been instrumental in helping cyber defenders secure critical software. It has also shown promise in life sciences research, generating novel scientific hypotheses. Anthropic intends to expand access to Mythos 5 through a broader trusted access program in the future.