Claude Fable 5.1: is the useful upgrade better work at lower effort?

Claude’s September 1 release gives users a more useful question than “which model tops the chart?”: can the same job now be completed with less correction, waiting or cost?

Anthropic introduced Fable 5.1 for general availability and Mythos 5.1 through restricted trusted-access programmes. They share an underlying model but have different safeguards. Mythos is not simply an upgrade everyone can select. Anthropic also reports improved performance at lower effort levels and estimates lower costs for typical token-billed workloads through cheaper cache reads. That is not a promise that every Claude subscription is cheaper or every task will use fewer tokens.

For a designer, writer or small team, a sensible test is to repeat one real brief: the same inputs, the same constraints and the same definition of finished. Try a lower effort setting first where available. Record whether the output answers the whole brief, how much editing it needs and what you have to check. A faster first draft that takes longer to repair has not saved much.

The official breakdown is worth reading for the distinction between capability, access and cost: [Anthropic’s Fable and Mythos 5.1 announcement](https://www.anthropic.com/claude-fable-and-mythos-5-1).

These are Anthropic’s release claims, not WhatAI benchmark results. We would particularly like to hear about repeatable work rather than a single impressive answer. If you have tried Fable 5.1, which task needs less intervention now, and which still needs the same supervision?