Anthropic Releases Claude Fable 5.1 and Claude Mythos 5.1: 52.6% on Terminal-Bench-Science and 75% Cheaper Cache Reads

Anthropic has announced the release of Claude Fable 5.1 and Claude Mythos 5.1, representing a significant advancement in the company's AI model lineup. These releases showcase Anthropic's dual approach to AI deployment, offering both publicly accessible and restricted variants of the same underlying technology. Fable 5.1 is now generally available across multiple platforms, while Mythos 5.1 remains limited to vetted organizations through Project Glasswing, Anthropic's restricted access program.

The new Claude models demonstrate substantial improvements in capability and cost-efficiency. Fable 5.1 has achieved a 52.6% score on Terminal-Bench-Science, indicating enhanced performance on complex scientific reasoning tasks. Most notably, the update introduces a 75% reduction in cache read costs, directly addressing operational expenses for high-volume API users. Fable 5.1 is accessible through multiple deployment channels including the Claude API, Amazon Web Services (AWS), Google Cloud Platform, and Microsoft Foundry, ensuring widespread availability for enterprise and developer communities. Mythos 5.1, meanwhile, maintains restricted access through Project Glasswing for organizations that meet Anthropic's vetting criteria.

The release of these models carries several important considerations for the AI sector:

• Significant cost reduction in cache operations could accelerate adoption of Claude among price-sensitive enterprise customers
• Dual-model strategy allows Anthropic to balance public accessibility with controlled deployment for sensitive applications
• Improved benchmark performance on scientific tasks positions Claude competitively against rival models in specialized domains
• Expanded availability across major cloud providers increases market penetration and developer adoption
• Project Glasswing's continued restriction suggests Anthropic's commitment to safety-first deployment practices

These releases underscore Anthropic's strategy of combining performance advancement with responsible AI deployment. The dramatic cost improvements make Claude more economically viable for large-scale applications, while the dual-variant approach demonstrates a thoughtful balance between innovation and safety. As competition intensifies in the large language model market, such improvements in both performance metrics and operational economics become critical differentiators in enterprise AI adoption decisions.