Anthropic Releases Mythos Open Version Without Cybersecurity Features

Jin10 Data, June 10—Anthropic launched the open version of its Mythos model on Tuesday local time, but implemented restrictions prohibiting its use in high-risk fields such as cybersecurity. Anthropic stated that the new Claude Fable 5 is the most powerful model they have developed so far for broader applications, praising its performance in software engineering and data analysis. Previously, Anthropic announced that Mythos had identified thousands of software vulnerabilities. Subsequently, through the "Glasswing" initiative, the company restricted the model's usage to approximately 200 institutions, including the U.S. government. Anthropic emphasized that extensive testing has been conducted to ensure users cannot manipulate Claude Fable 5 to bypass its guidelines and perform restricted operations. Dianne Penn, Anthropic's Head of Product Management, Research, and Labs, noted that based on early customer feedback, while Fable 5 is priced higher, it consumes fewer tokens to execute tasks, thereby reducing the overall cost per task. Anthropic also stated that users with access to the Claude Mythos preview version (i.e., the version without security protection mechanisms) will be able to upgrade to the new Claude Mythos 5. Both models are priced at $10 per million input tokens and $50 per million output tokens.