Anthropic launches Claude Fable 5.1 and says it’s up to 45 percent cheaper for agentic work

Anthropic’s latest update is aimed as much at annoyed customers as at benchmark bragging rights. The company says Fable 5.1 and Mythos 5.1 were built to answer complaints about three things people kept bringing up: cost, data retention, and safeguards that got a little too eager.

The headline number is the one Anthropic clearly wants people to remember. Claude Fable 5.1, it says, performs better than Fable 5 while usually costing about 25 percent less. For more complex agentic work, the savings can reach up to 45 percent. The company says that’s partly because it cut pricing on cached data — the stuff that was already processed and stored.

That makes this more than a routine model refresh. Anthropic is trying to make the case that its models can be both more capable and less expensive at the same time, especially for the kind of tasks where AI agents chew through a lot of reused context. That’s a practical promise, not just a shiny one.

Early reactions started surfacing alongside the announcement. Every CEO Dan Shipper called it the strongest coding model his team has used, while also saying it’s faster, more token-efficient, and sounds more natural. That kind of feedback won’t settle the debate, but it does suggest Anthropic has landed on a release people are ready to poke at immediately.