Anthropic released Claude Fable 5 — the public version of the Mythos-class model, the most powerful from the company to date 🏋️‍♂️

A strong leap in coding, working with large context, documents, and visual tasks: images, tables, graphs, screenshots.

Since April, the model has been in closed preview. In tests, it independently found previously unknown vulnerabilities in popular operating systems and browsers and could assemble a working exploit from them. Access was only available to partners (AWS, Microsoft, Apple, CrowdStrike) under the Project Glasswing program.

💬 A separate point — built-in protection against dangerous scenarios. Requests related to cybersecurity, biology, chemistry, and model distillation are automatically switched to Opus 4.8. According to Anthropic, filtering triggers in less than 5% of sessions.

In subscriptions Claude Pro, Max, Team, and Enterprise, the model is free until June 22, after that — via credits or API.

API price: $10 per 1 million input tokens and $50 per 1 million output tokens, twice as expensive as Opus 4.8 🤑

Video: claudeai

🌐 Website | 🎰LinkedIn