Anthropic has expanded access to its advanced AI models by making Claude Fable 5 publicly available and simultaneously launching Claude Mythos 5 for a limited number of partners as part of the Project Glasswing collaboration initiative with the US government.

Both models are built on a common basic architecture. The key difference is the level of protection limits: Mythos 5 operates with weakened standard filters. Earlier in April, access to the pre-release version of Mythos was granted to a narrow range of cybersecurity specialists and critical software infrastructure providers.

Claude Fable 5 is now available to all users via the API and consumer-pay corporate plans. Until June 22, 2026, using the model does not require additional credits in Pro, Max, Team and seat-based Enterprise subscriptions. From 23 June, use credits will be required to work with her if the company does not extend the grace period. To implement security in sensitive areas (cybersecurity, biology, chemistry, model distillation), Fable 5 automatically redirects requests to the more powerful Claude Opus 4.8 - according to Anthropic, such a fallback works on average in less than 5% of sessions.

The cost of using both models is $10 per 1 million input tokens and $50 per 1 million weekend. In the future, Anthropic plans to expand access to Mythos 5 for individual biologists and trusted-access program members.

Recall that in May, Anthropic released Claude Opus 4.8 and separately introduced to Claude Code a dynamic workflow function.