New York: Anthropic has launched Claude Fable 5.1 and Claude Mythos 5.1, its latest AI models aimed at coding, knowledge work, cybersecurity and scientific research. The company says the new models deliver stronger performance on complex and long-running tasks while reducing the cost of running agentic workloads.
Fable 5.1 is available to Pro, Max, Team and Enterprise customers, as well as developers through Anthropic’s Claude Platform and cloud partners. Mythos 5.1 has more restricted access and is being offered to vetted organizations working in areas including cybersecurity and life sciences.
The two models use the same underlying system but operate with different safeguards. Anthropic has designed Fable 5.1 for broader use, while Mythos 5.1 retains additional controls because of its capabilities in areas that could present greater risks if misused.
Pricing is also changing with the new release. Fable 5.1 costs $10 per million input tokens and $50 per million output tokens. Anthropic has reduced cache-read pricing to $0.25 per million tokens, which the company estimates will lower typical workload costs by about 25 percent and highly agentic workloads by up to approximately 45 percent compared with Fable 5.
Anthropic says Fable 5.1 is designed to handle projects that can run for hours or span multiple applications. The model can plan tasks, use tools, recover from failed steps and continue working with limited human intervention. Its coding capabilities are aimed at large codebases, code reviews, performance improvements and extended autonomous coding sessions.
The company is also highlighting gains in scientific and technical work. Fable 5.1 scored 52.6 percent on Terminal-Bench-Science 0.1, compared with 24.7 percent for Fable 5 in Anthropic’s testing. The company says the model is intended to support increasingly complex research and knowledge-based tasks.
Safety controls have also been adjusted. Anthropic says Fable 5.1 can identify software vulnerabilities in source code, while protections continue to block penetration testing, exploit generation and binary-based vulnerability scanning. The company also says its biology safeguards now intervene on benign requests 85 percent less often than the safeguards introduced with Fable 5.
The changes come shortly after Anthropic disclosed problems identified during cybersecurity evaluations of earlier models. In a separate August 31 update, the company discussed improvements to its alignment and security practices following incidents involving models accessing real-world systems during testing.
Anthropic is also working on Enterprise Frontier Safeguards, which are intended to give enterprise customers greater control over where their data is stored while maintaining protections against adversarial use. The company has separately introduced invisible text watermarking for newer Claude models as part of its efforts to identify AI-generated content.
With Fable 5.1 now generally available and Mythos 5.1 limited to vetted organizations, Anthropic is expanding access to its most capable model while keeping stronger controls around higher-risk cybersecurity and biology capabilities.
Previous
