Anthropic has announced the launch of Claude Fable 5, describing it as its most capable AI model released for general use to date. The company claims the new model sets a new benchmark across software engineering, scientific research, knowledge work, vision and complex reasoning tasks, while introducing additional safeguards designed to mitigate potential misuse.

According to Anthropic, Claude Fable 5 outperforms its predecessors on nearly all evaluated benchmarks and demonstrates a growing advantage as tasks become longer and more complex. The company highlighted the model's ability to work autonomously on extended projects, enabling significant productivity gains across enterprise and research environments.

Recognising the risks associated with increasingly powerful AI systems, Anthropic has implemented a safety framework that automatically redirects certain high-risk queries, particularly those related to cybersecurity, to its less capable Claude Opus 4.8 model. The company stated that the safeguards were intentionally configured conservatively to enable a faster deployment while minimising the risk of misuse.

Alongside Fable 5, Anthropic introduced Claude Mythos 5, a version of the same underlying model with reduced restrictions in selected cybersecurity domains. Initially, access to Mythos 5 will be limited to approved cyber defence organisations and infrastructure providers through Project Glasswing, a collaboration with the U.S. government. Anthropic claims Mythos 5 offers industry-leading cybersecurity capabilities and plans to expand access through a broader trusted-user program.

The company also announced a significant reduction in pricing, with both models available at $10 per million input tokens and $50 per million output tokens, less than half the cost of the earlier Claude Mythos Preview.

In software engineering applications, Anthropic cited early testing results from Stripe, where Fable 5 reportedly completed a large-scale code migration involving a 50-million-line Ruby codebase in a single day, a task estimated to require more than two months of manual work by an engineering team.

Anthropic said the new models are already demonstrating potential benefits in cybersecurity, scientific discovery and life sciences research, where they are helping researchers generate novel hypotheses and accelerate therapeutic development. The launch reflects the company's broader strategy of delivering increasingly advanced AI capabilities while balancing safety, accessibility and real-world impact.