Anthropic has released Fable 5.1, citing Improved Programming Capability and Cost Reduction

Perceive Beating AI Newsletter: Anthropic Launches New Version of AI Model Fable 5.1, Touting Improved Performance in Programming and Scientific Tasks with Cost Efficiency.

In the programming realm, Fable 5.1 excels in handling long-term, complex tasks, including software project development, code reviews, and intricate scenarios involving entire application code. Its scientific capabilities have also been enhanced, covering experimental design, simulation of experimental results, and parsing of complex charts and data tables.

On the cost front, enterprise users' per-token pricing remains consistent with Fable 5, but a 75% reduction has been applied to reinvoking parts of the model that have processed information. This segment may constitute over half of the token usage in long texts and complex tasks, with the anticipated significant reduction in overall operating costs post-reduction.

Regarding security, Fable 5.1 has upgraded its classifier, enhancing the accuracy of risk instruction identification, reducing false positives, enabling more efficient utilization by users in cybersecurity and biology-related fields. Moreover, Anthropic plans to allow enterprise customers to store data from the most robust models on their proprietary cloud infrastructure rather than on its servers, while continuing security checks.