Anthropic has released Claude Fable 5, its first publicly available Mythos-class AI model. The company previously deemed its Claude Mythos model too powerful for general release, limiting access to select partners like Apple for cybersecurity applications. Now Fable 5 brings advanced capabilities to paying customers while incorporating safety measures to prevent misuse.

Fable 5 delivers state-of-the-art performance across software engineering, knowledge work, vision, and scientific research tasks. Anthropic claims the model's lead over previous versions grows with task complexity, enabling longer autonomous operation than any prior Claude model. Benchmark results reportedly favorably compare against Claude Opus 4.8 and GPT-5.5.

Pricing runs $10 per million input tokens and $50 per million output tokens, undercutting the original Claude Mythos Preview costs. Through June 22, Fable 5 comes free with Pro, Max, Team, and Enterprise subscriptions. After that date, usage requires credits until capacity allows broader availability.

The release addresses growing demand for powerful AI while balancing safety concerns. Anthropic's conservative safeguard approach routes sensitive queries to less capable models, though this catches some legitimate requests. The company plans to refine these protections as more advanced models arrive.