Anthropic Announces Claude Fable 5.1 and Mythos 5.1
New AI models reduce cache read costs by 75%.
1Anthropic's New Models
Anthropic has unveiled the Claude Fable 5.1 and Mythos 5.1 models. These models offer the same technology with different safety measures. Cache read prices have been reduced by 75%, dropping to $0.25 per million tokens. This cuts typical workload costs by 25% and agentic workloads by up to 45%. The models show significant performance improvements in scientific research benchmarks compared to their predecessors.
2Performance Boost
Fable 5.1 scored 52.6% on Terminal-Bench-Science 0.1, while the previous version, Fable 5, scored only 24.7%. On Terminal-Bench 4.0, it achieved an even higher score of 55.8%. This improvement is based on Anthropic's own data. The models were developed with a focus on safe AI systems. The company was founded by former OpenAI members and specializes in AI safety.
3EU Legal Compliance
Anthropic signed the EU's Code of Practice on the transparency of AI-generated content in July. Among over 190 signatories, the company now marks Claude outputs with a watermark. Applicable to models released after August 2, 2026, this watermark is invisible without a detection tool. It reportedly carries no user information. This move ensures compliance with EU regulations.
4Detection API
Anthropic plans to offer a free detection API. Currently in private preview, this tool is only available to eligible organizations. These include regulators, law enforcement, media, fact-checkers, independent researchers, educational institutions, and EU civil society groups. The company aims to expand access over time but has not provided a specific timeline.
5Mythos Access Programs
The Mythos model is available only to vetted users through two access programs covering cybersecurity and life sciences. Developed in collaboration with the US government, the biology program is currently open only to select US organizations. Anthropic is working to include international partners but European cyber defenders and life scientists are not yet part of the program.
6Legal Requirements
Article 50 of the EU's regulations requires providers to mark synthetic outputs in a machine-readable format and make them detectable. This rule has been in effect since August 2, 2026. Enterprises with their own compliance obligations can also apply. Anthropic plans to expand access over time but has not shared a specific timeline.
7Balancing Technology and Security
Anthropic's models aim to strike a balance between technology and security. While Fable 5.1 is generally available, Mythos is offered to select users with stricter safety measures. The company employs various methods, such as watermarks, to ensure the safe use of AI in compliance with EU laws. This approach reflects its commitment to responsible AI development.
8Industry Impact
Anthropic's new models are seen as a significant step in the AI industry. Reduced costs and improved performance enhance the company's competitiveness. Compliance measures with EU regulations set an example for other AI firms. Mythos's limited access underscores its security-focused approach. These developments provide important insights into the future of AI.
Original article: The Next Web — Anthropic, Claude Fable 5.1 ve Mythos 5.1'i Duyurdu