Signal

Claude Fable 5.2 and Mythos 5.2

First reported by Anthropic ·

The signal ●●●● Compiled by AI from Anthropic, Hacker News, TechCrunch, The Verge, The Register and 19 more
Why you might care

Claude Fable 5.2 and Mythos 5.2 are now up to 30% faster and 40% cheaper for complex tasks, potentially lowering the cost of advanced AI-assisted development.

What happened

Anthropic has released Claude Fable 5.2 and Claude Mythos 5.2, updated AI models designed for complex technical challenges. These new versions demonstrate enhanced capabilities in coding, reasoning, and vision, with prerelease testing showing significant achievements like automating software migrations, designing electronic devices from scratch, and contributing to physics research. Fable and Mythos are the same model, with Fable incorporating stronger safeguards for broader availability, while Mythos is accessible to organizations in cybersecurity and biomedical research via specialized programs. The new models offer a communication improvement and are up to 30% faster than their predecessors, Claude Fable 5.1 and Claude Mythos 5.1. They are also priced lower, at $8 per million input tokens and $40 per million output tokens, with an optional fast mode available at double these rates. Anthropic has implemented enhanced safety testing and protections, including against distillation attacks, to keep pace with the models' growing abilities.

What it means

The introduction of Claude Fable 5.2 and Mythos 5.2 signifies Anthropic's push to make highly capable AI more accessible and cost-effective for demanding technical work. By offering these models at a lower price point and with increased speed, Anthropic is enabling a broader range of organizations and individuals to tackle complex problems in areas like scientific research, cybersecurity, and advanced software engineering. The distinction between Fable and Mythos, with the latter offering specialized access for sensitive fields, also indicates a tiered strategy for deploying powerful AI responsibly.

These releases suggest a market trend toward optimizing large language models not just for raw capability, but also for efficiency and cost. The benchmark data indicates Fable 5.2 and Mythos 5.2 outperform or match top-tier competitors like GPT-6 Astra and Gemini 4 Carbon on various metrics, often at a fraction of the cost. This focus on performance-per-dollar could drive increased adoption of advanced AI for enterprise-level solutions, potentially reshaping the competitive landscape for AI development platforms and services.

AI-written summary. May contain errors.

Claude