Anthropic launched a public version of its Mythos model but prohibits its use in cybersecurity and other high-risk areas. The company also introduced Claude Fable 5, which it called its most capable model for broader use, citing strong performance in software engineering and data analysis. Anthropic said Mythos previously uncovered thousands of software vulnerabilities and had earlier restricted access to about 200 institutions, including the US government, under a Glasswing program. It said it

Anthropic launched a public version of its Mythos model but prohibits its use in cybersecurity and other high-risk areas. The company also introduced Claude Fable 5, which it called its most capable model for broader use, citing strong performance in software engineering and data analysis. Anthropic said Mythos previously uncovered thousands of software vulnerabilities and had earlier restricted access to about 200 institutions, including the US government, under a Glasswing program. It said it conducted extensive testing to prevent users from manipulating Claude Fable 5 to bypass safeguards. Early customers reported Fable 5 carries a higher price per unit but uses fewer tokens per task, lowering total task cost. Users with access to the unprotected Claude Mythos preview can upgrade to Claude Mythos 5. Both models are priced at $10 per mln input tokens and $50 per mln output tokens.