Anthropic unveils Claude Fable 5.1 and Mythos 5.1 with sharper scientific focus
The AI firm’s latest models promise a 25 per cent cost reduction for typical workloads while introducing new safeguards for enterprise data privacy and advanced capabilities in fields ranging from protein design to planetary mapping.
Anthropic has launched Claude Fable 5.1 and Claude Mythos 5.1, positioning the pair as the world’s most advanced models for coding and knowledge work. While both models share the same underlying architecture, they differ in their availability and safeguard levels. Claude Fable 5.1 is generally available to the public, whereas Claude Mythos 5.1 is restricted to trusted access programs designed specifically for cybersecurity and life sciences. The release marks a significant step in the company’s strategy to address customer feedback regarding pricing, data retention, and the precision of safety filters.
For enterprise users, the financial implications of the update are notable. Anthropic has reduced the cost of cache reads by 75 per cent, bringing the price down to $0.25 per million tokens. This adjustment results in an estimated 25 per cent reduction in costs for typical workloads compared to the previous Fable 5 model. For highly agentic tasks, which rely heavily on context reuse, the savings could reach up to 45 per cent. The base pricing remains unchanged at $10 per million input tokens and $50 per million output tokens, but the structural shift in cache costs aims to make the model more economical for sustained, complex operations.
A key feature of the launch is the introduction of Enterprise Frontier Safeguards, or EFS. This system allows enterprise customers to store data on their own cloud infrastructure rather than on Anthropic’s systems, effectively providing the privacy of a zero data retention policy. Developed in collaboration with more than 100 customers and major cloud partners including Amazon Web Services, Google Cloud, and Microsoft Azure, EFS is designed to prevent adversarial use while maintaining data sovereignty. The rollout will begin in phases later this fall, with eligible customers able to use Fable 5.1 with zero data retention in the interim.
Beyond cost and privacy, the models demonstrate enhanced capabilities in scientific research. Claude Mythos 5.1 has shown particular promise in molecular design, creating protein binders with affinities 10 times higher than the best designs submitted to Adaptyv Bio’s competitions. In computational biology, the model sped up seven open-source deep learning models by up to 2.5 times through the creation of custom GPU kernels, reducing estimated GPU costs by 30 to 60 per cent. Additionally, Claude Fable 5.1 generated a high-resolution elevation map of one-third of the planet Venus, revealing details down to two to three kilometres based on radar images from NASA’s Magellan mission.
The launch also addresses previous criticisms of safety filters that often flagged benign content. Anthropic reports that its updated cybersecurity safeguards for Fable 5.1 block 60 per cent fewer false positives than previous versions. This allows the model to be used for discovering software vulnerabilities without necessarily developing exploits, a distinction that is crucial for defensive cybersecurity work. In the life sciences sector, a new access program developed in partnership with the US government will enable scientists to utilise the advanced biology capabilities of Claude Mythos 5.1.
The models are available immediately on major cloud platforms, including Amazon Web Services, Google Cloud, and Microsoft Azure. Anthropic has also introduced an anti-distillation mechanism to prevent new API accounts from extracting the model’s thinking through manual context editing. Furthermore, in line with the EU AI Act, a compliance watermark has been added to model outputs released after 2 August 2026, with a detection API currently in private preview for eligible organisations.
