Fable 5 is Anthropic’s newest AI model for simulating human behavior and building AI agents. It can copy personality traits, decision-making styles, and complex interactions much more naturally than previous models.
Why it matters: companies can create digital versions of employees, customers, or groups of users and test decisions before applying them in the real world.
Some interesting facts:
It was trained on behavioral and psychological data, not just text.
It is designed for market research, product testing, and organizational simulations.
It can create thousands of virtual personas with different behaviors.
It is one of the most ambitious attempts to build large-scale “digital twins” of humans.
It is the most powerful model Anthropic has released publicly so far and ranks among the best AI models for coding, long-form analysis, and multi-step tasks.
It has a 1 million token context window, allowing it to process huge amounts of text, files, and information at once.
The public version is actually a restricted version of the more powerful Mythos 5 model and includes extra limitations in sensitive areas such as cybersecurity.
Some companies report that Fable 5 can work independently on a project for hours, bringing AI one step closer to fully autonomous agents.
API Pricing:
$10 per million input tokens
$50 per million output tokens
That makes it roughly twice as expensive as Opus 4.8 and one of the most expensive commercial AI models available.
One of the most interesting facts is that only a few days after its release, the U.S. government restricted access to Fable 5 and Mythos 5 for some foreign users due to security concerns—something that has happened to very few AI models at this level.