Anthropic Releases Claude Fable 5.1 and Mythos 5.1 with Significant Price Cuts and Enhanced Safety Features
Anthropic has announced the release of Claude Fable 5.1 and Mythos 5.1, cutting cache read prices by a staggering 75%. These new models, based on the same architecture but featuring distinct safeguards, demonstrate impressive performance: Fable 5.1 achieves 52.6% on a scientific research benchmark, surpassing its predecessor’s 24.7%.
Key Highlights:
- Price Reduction: Cache read costs have been lowered to $0.25 per million tokens, resulting in significant cost savings for users, especially for workload and agentic applications (up to 45%).
- Benchmark Performance: Fable 5.1 scores 52.6% on Terminal-Bench-Science 0.1 and 55.8% on Terminal-Bench 4.0, marking a substantial improvement.
- AI Safety Features:
- The models are compliant with the EU AI Act, featuring a watermark on outputs to ensure transparency.
- A detection API, currently in private preview for eligible organizations (regulators, media, researchers, and EU civil society groups), enables verification of synthetic content.
- Limited Availability: Mythos is accessible through two access programs, one focused on cybersecurity and another on life sciences, with partnerships involving the US government.
In a move towards wider adoption, Anthropic aims to expand API access over time. The company’s actions align with European regulations, ensuring that AI-generated content is properly marked and detectable, as required by Article 50 of the EU AI Act.