Anthropic has unveiled significant updates to its Claude AI model lineup with the release of Claude Fable 5.1 and Claude Mythos 5.1. These new versions come with enhanced performance, improved efficiency, and a shift in access policies. Fable 5.1 is now generally available across major cloud platforms, including the Claude API, AWS, Google Cloud, and Microsoft Foundry, while Mythos 5.1 remains restricted to select organizations under the Project Glasswing initiative.
Performance and Efficiency Gains
One of the most notable improvements in Fable 5.1 is its performance on scientific benchmarks. The model achieved a score of 52.6% on Terminal-Bench-Science 0.1, a marked improvement over Fable 5's score of 24.7%. This advancement highlights Anthropic's ongoing efforts to refine the reasoning and analytical capabilities of its AI models. Additionally, the new version supports a context window of 1 million tokens, enabling it to handle more complex and extended inputs.
Another key enhancement is the significant reduction in cache read costs. Anthropic has slashed cache read pricing by 75%, now charging just $0.25 per million tokens, compared to the previous rate of $1.00. While the base pricing for tokens remains unchanged at $10 and $50 for input and output respectively, this reduction is expected to make the model more cost-effective for developers and enterprises running intensive AI workloads.
Breaking API Changes
Anthropic also announced several breaking changes to its API, including the removal of forced tool use. This change allows developers more flexibility in how they integrate and use the Claude models, potentially streamlining workflows and reducing friction in development processes.
Overall, the release of Claude Fable 5.1 and Mythos 5.1 underscores Anthropic’s commitment to advancing both the performance and accessibility of its AI tools, while also refining its pricing and API structure to better serve its growing user base.
