As we initially covered yesterday, after some heavy lobbying by Anthropic, U.S. Commerce Secretary Howard Lutnick lifted the restrictions on Fable 5 and Mythos 5 after fears that they could be too powerful.
The following day, Thariq Shihipar, a member of Claude's technical staff at Anthropic, and Cat Wu, the company's head of product for Claude, held a roundtable hosted by Simon Willison, founder of the Datasette open-source project, to promote the new model and sing the praises of its efficacy.
The Anthropic duo claimed that Fable blew past competitor models in terms of speed and efficiency, with agentic coding and code fixing particular strengths. In the latter case, it was dealing with 50% more pull requests for its engineers.
The complexity of the code that Fable can create is better than anything Anthropic has managed before, the team asserted. It can operate for longer without needing human input, check its own prompts by writing test scripts for software it has created, and handle visual information from embedded graphs and images in source material.
"It's good at the dynamic permissions that you yourself give inside of the prompt, which I think is really important," Shihipar said.
