Anthropic Launches Public Version of Claude Fable 5 Mythos-class Model
Anthropic released Claude Fable 5 on Tuesday local time, which is the secure public version of its strongest Mythos-class model, prohibited for use in high-risk areas such as cybersecurity.
The model excels in software engineering and data analysis, having discovered thousands of software vulnerabilities through Mythos; Anthropic restricts unprotected Mythos to about 200 institutions (including the U.S. government) through the Glasswing program.
The new version is priced at $10 per million input tokens and $50 for output. Product lead Dianne Penn stated that although the unit price is higher, the token consumption for tasks has significantly decreased, resulting in overall cost efficiency; preview users can upgrade to Claude Mythos 5.
Source: Public Information
ABAB AI Insight
Anthropic previously limited Mythos Preview to a few defensive partners through Project Glasswing. The release of Fable 5 continues its path of "capability leap with strong safeguards," having discovered thousands of zero-day vulnerabilities in Mythos Preview testing and choosing not to publicly release them to prevent misuse.
On the capital path, Anthropic mobilizes internal research resources and collaborates with enterprises/cloud platforms to transform Mythos-class capabilities into widespread subscription revenue through Fable 5, while maintaining a high-priced restricted channel for Mythos 5, creating a dual monetization strategy that expands the user base while preserving the premium for high-end security services.
Similar to OpenAI balancing capability and safety through the GPT series iterations, Anthropic is currently in an expansion phase transitioning from cutting-edge closed models to tiered commercialization, thereby solidifying its position as a leader in responsible AI.
Essentially, this reflects technological substitution and regulatory changes: Fable 5 democratizes Mythos capabilities through safeguards, directly replacing traditional coding/analysis toolchains, while regulatory restrictions accelerate the shift of AI capabilities from indiscriminate openness to tiered control, reshaping the industrial chain structure of high-risk technologies in software security and enterprise productivity.