Anthropic has launched Claude Fable 5.1 and Claude Mythos 5.1, two new AI models designed to improve coding, research, knowledge work, and long-running agentic tasks while lowering the cost of demanding AI workloads.
Claude Fable 5.1 is the version available to general customers, while Claude Mythos 5.1 is aimed at approved organizations and researchers working in areas such as cybersecurity and life sciences. Both models are based on the same underlying technology, but Mythos offers different safeguards for vetted professional use.
One of the biggest changes in Fable 5.1 is lower pricing for cached data. AI agents often need to repeatedly read the same instructions, documents, code, tools, and conversation history while completing a long task. Instead of processing all of that information from scratch every time, Claude can reuse previously processed context through caching.
Anthropic has reduced the price of cache reads by 75%, bringing the cost down to $0.25 per million tokens. The company estimates that this could make typical Fable 5.1 workloads around 25% cheaper than Fable 5.
The savings can become even larger for agentic AI systems. For complex coding projects and other tasks where an AI agent repeatedly works with large amounts of existing context, Anthropic says costs could fall by up to around 45%.
The standard API pricing remains $10 per million input tokens and $50 per million output tokens. This means the main cost improvement comes from making repeated access to previously processed information much cheaper.
Fable 5.1 also brings performance improvements for coding, knowledge work, computer-based tasks, scientific research, and complex problem solving. Anthropic says the model is better at working through long tasks instead of simply producing a quick answer.
This is especially important for AI agents that may need to inspect files, use tools, write and test code, investigate problems, and continue working through several stages before completing a job.
Anthropic has also adjusted its safety systems to reduce cases where harmless requests are incorrectly blocked. In cybersecurity work, the company says its updated safeguards result in around 60% fewer interventions compared with the previous Fable model.
Fable 5.1 can now help identify software vulnerabilities for defensive security work. More sensitive activities, including exploit generation and some penetration-testing tasks, remain more restricted.
The company is also introducing Enterprise Frontier Safeguards, or EFS, for organizations that need stronger privacy controls. The system is designed to allow customer data to remain inside cloud infrastructure controlled by the customer while still providing protections against misuse. The feature is expected to roll out to enterprise customers in phases.
Claude Mythos 5.1 takes a different approach. It is designed for vetted cybersecurity and life-sciences users who need access to advanced capabilities that would normally face tighter restrictions in the standard Claude experience. Access will therefore remain limited rather than being opened to everyone.
For businesses, the combination of stronger agentic capabilities and cheaper context reuse could be particularly important. AI agents increasingly need to work with large codebases, company documents, databases, tools, and long-running conversations. Reducing the cost of repeatedly accessing that information could make advanced AI automation more practical for everyday enterprise use.
Claude Fable 5.1 is available through Anthropic’s platforms and API as well as supported cloud services, including Amazon Web Services, Google Cloud, and Microsoft’s cloud ecosystem.
With Fable 5.1 and Mythos 5.1, Anthropic is focusing not only on making Claude smarter, but also on making advanced AI agents more economical, useful, and manageable for organizations running complex workloads.
