Claude Fable 5 and Claude Mythos 5: A Revolutionary Leap in AI Capabilities and Safety
Anthropic, a leading AI research company, has unveiled Claude Fable 5 and Claude Mythos 5, marking a significant milestone in the evolution of AI technology. These models represent a leap forward in AI capabilities, offering state-of-the-art performance in various domains, while also addressing critical safety concerns.
Unparalleled Capabilities
Fable 5 showcases exceptional capabilities across multiple domains, surpassing previous models in software engineering, knowledge work, vision, scientific research, and more. It excels in complex tasks, compressing months of engineering into days and demonstrating superior token efficiency. In knowledge work, Fable 5 outperforms peers on finance benchmarks, showcasing advanced reasoning and document interpretation.
In the realm of vision, Fable 5 sets a new standard, extracting precise numbers from scientific figures and performing complex tasks like rebuilding source code from screenshots. Its memory and long-context capabilities are transformative, enabling it to stay focused and improve outputs using its own notes.
Drug Design and Scientific Hypotheses
Mythos 5, a variant of Fable 5, accelerates drug design processes by tenfold, matching or surpassing skilled human operators. It consistently generates novel, compelling scientific hypotheses in molecular biology, with one hypothesis corroborated by an independent study. Mythos 5's autonomous research capabilities are remarkable, assembling single-cell data and designing machine learning models, outperforming existing models in genomics research.
Safety and Alignment
Anthropic prioritizes safety, implementing robust safeguards in Fable 5. These safeguards include classifiers that detect potential misuse, jailbreak attempts, and prevent the model from responding on sensitive topics like cybersecurity, biology, and chemistry. The classifiers are designed to withstand sophisticated attempts to bypass them, ensuring the model's safe and controlled usage.
Trusted Access and Collaboration
Anthropic collaborates with the US government through Project Glasswing, providing Claude Mythos 5 to a select group of cyber defenders and infrastructure providers. This partnership aims to enhance cybersecurity capabilities and protect critical software. Additionally, a trusted access program for biology researchers will be introduced, allowing access to Fable 5 with specific safeguards removed.
Pricing and Availability
Fable 5 and Mythos 5 are offered at competitive pricing, with $10 per million input tokens and $50 per million output tokens. Developers can access Fable 5 via the Claude API, with subscription plans available in stages. Mythos 5, with cyber safeguards lifted, is initially restricted to Project Glasswing partners and will expand through a trusted access program.
Conclusion
Anthropic's release of Claude Fable 5 and Claude Mythos 5 represents a significant advancement in AI technology, combining unprecedented capabilities with a strong emphasis on safety. These models have the potential to revolutionize various industries, from software engineering to drug discovery, while also fostering collaboration and innovation in the AI community.
