Anthropic’s New Model Targets Power Users But Cuts Off AI Rivals

App developers finally got their first real look at Anthropic’s Mythos Tuesday when Anthropic released a somewhat neutered version of the model, Claude Fable 5, following weeks of breathless chatter about it.

Developers who tried the new model said the AI largely deserved the hype. They said it thoroughly outperformed earlier Anthropic models in coding, general knowledge work and spatial reasoning, although questions touching on cybersecurity are routed to the weaker Claude Opus 4.8 AI to prevent misuse.

But early testers said that it was hard to notice much of an improvement over Anthropic’s past models on easy or medium-difficulty tasks, including the typical questions you might ask Google Search or ChatGPT. Where Fable really stood out, they said, was on long-duration, complex tasks, like building an entire video game with just a single prompt.