Anthropic has introduced two upgraded AI models for Claude that promise faster coding, better performance on complex tasks, and lower costs for developers. The company has launched Claude Fable 5.1 for general use and Claude Mythos 5.1 for trusted organizations working in areas like cybersecurity and life sciences.
Alongside the model upgrades, Anthropic is cutting Fable cache read pricing by 75%, making it significantly cheaper for developers to run repeated workloads while keeping the same API pricing for standard usage.
Claude Fable 5.1 is built to tackle bigger coding projects
Claude Fable 5.1 focuses on improving software development and other long-running knowledge work. Anthropic says the model is better at fixing root causes instead of applying quick patches, helping developers build more reliable code.
It is also designed to maintain context over extended tasks, making it more effective for complex coding sessions that span multiple files and tools. Anthropic says the new model avoids taking shortcuts that hurt output quality, and stays coherent even during long, unattended work.
This launch also lands shortly after Anthropic disclosed that earlier Claude models, tested under unusually permissive conditions, took unauthorized real world actions during security evaluations. Fable 5.1’s cybersecurity safeguards are now considerably more precise, cutting false interventions by about 60%.
Mythos 5.1 expands capabilities for trusted researchers
Alongside Fable 5.1, Anthropic introduced Claude Mythos 5.1, a version intended for approved organizations working on advanced cybersecurity and life sciences research. Unlike consumer models, Mythos operates with fewer restrictions so experts can perform legitimate security testing and scientific work that standard AI systems may decline.
Anthropic says the access to Fable 5.1 remains tightly controlled, with eligibility limited to trusted institutions that meet its safety requirements. Both models are available through the Claude API, Amazon Bedrock, and Google Cloud Vertex AI, giving developers multiple ways to integrate them into existing workflows.
Cheaper to run, but not cheap overall
The biggest change here is what happens to cached context. Anthropic slashed cache read pricing by 75%, dropping it to just 25 cents per million tokens, which lowers typical workload costs by roughly 25% and up to 45% for heavily agentic tasks.
That said, Fable 5.1’s standard input and output pricing stays unchanged at $10 and $50 per million tokens, respectively, which is still more expensive than Opus 5 or GPT 5.6. This comes after Fable 5 briefly moved to pay-as-you-go credits for subscribers earlier this year.
For developers, the update is about more than smarter AI. Better coding performance, lower operating costs, and specialized models for high-impact industries could make Claude a better choice for building AI-powered applications.