Anthropic Says New Fable AI Model Is Cheaper, Better at Coding

(Bloomberg) -- Anthropic PBC is releasing a new version of its powerful Fable artificial intelligence model that it says is better at coding and science tasks, as well as more economical — all changes it says are in response to customer feedback.

The company said Fable 5.1, which it plans to release on Tuesday, is more skilled at carrying out lengthy, complicated computer programming tasks, such as software projects or code reviews that require it to work across all the code for an application. The new version is also better at helping with scientific work like coming up with experiment designs, simulating what could happen during an experiment and reading information in dense diagrams and tables, Anthropic said.

Companies are increasingly trying to justify and rein in the costs for AI usage at a time when open-weight, lower cost models from China are rapidly improving. These trends are applying pressure to US-based model makers like Anthropic.

Anthropic will charge enterprise users the same price per token — a unit of data processed by AI models — for Fable 5.1 as it has for Fable 5, but is lowering the price it charges users when the AI model resurfaces information it had previously processed.

Anthropic said the 75% price reduction it's offering on this part of the process will lead to a "substantial reduction" in the total cost of running the model.

The process can account for half or more of the tokens used to carry out some of the lengthier, more complicated tasks people run on Anthropic's AI models, said Dianne Penn, who leads product management for research and labs at the company.

"This was one area we wanted to make improvements on, responding to customer feedback," she said.

The model is also meant to be better at carrying out tasks while using fewer tokens, Penn added.

Fable 5.1, the company said, has also improved safety classifiers — the software that spots and stops problematic commands given to a model. Now they are more accurate, with many fewer false positives, which means users working on cybersecurity and biology-related tasks will be able to do so more efficiently.

Anthropic added that it plans to allow business customers to retain data on their own cloud computing infrastructure when using its most capable artificial intelligence models, rather than on Anthropic's, while still running the company's safety checks on that data. It marks a stark change from an earlier data retention policy intended to mitigate potential cyber risks, confirming previous Bloomberg reporting.