Anthropic’s Claude Fable 5 is a version of Mythos the public can access today

Anthropic has officially opened access to its most powerful AI system yet, launching Claude Fable 5, a public-facing version of its previously restricted Mythos model. Announced on Tuesday, the rollout marks a major step in bringing frontier-level AI capabilities to everyday users, though with strict safety guardrails in place to prevent misuse in sensitive areas.

The company says Fable 5 is designed to excel in high-level tasks such as software engineering, knowledge work, and visual reasoning. However, it does not operate without limits. In high-risk domains like cybersecurity, biology, and chemistry, the system is programmed to block responses and instead redirect queries to a safer fallback model, Claude Opus 4.8.

Originally introduced in a controlled preview earlier in the year, the underlying Mythos system was only available to select partners due to safety concerns. It was later expanded to hundreds of organizations across multiple countries before this broader public release. Now, Fable 5 is accessible through Anthropic’s API and enterprise subscription plans, though its availability in consumer tiers will shift to a usage-credit model after June 23.

Anthropic’s rollout comes at a critical moment as global AI competition intensifies, with companies racing toward increasingly autonomous systems. The firm has also warned about the possibility of recursive self-improvement, where AI systems could eventually enhance themselves without human intervention. To reduce risks, Anthropic has implemented strict safety testing, including extensive jailbreak stress tests and a mandatory 30-day data retention policy for monitoring unusual behavior.

Early industry reactions suggest strong performance gains. Testing partners reported that Fable 5 outperforms competitors in complex analytical tasks, app development, and design workflows, while also delivering more consistent reasoning. However, the model’s high pricing—$10 per million input tokens and $50 per million output tokens—may limit mass adoption, especially among cost-sensitive enterprises already struggling with rising AI expenses.

source: techcrunch