Anthropic's Claude Gets Cheaper and Sharper With Fable 5.1

Anthropic has released Claude Fable 5.1 and Claude Mythos 5.1, the newest versions of its most capable models, with a pitch that leans as much on price as on raw ability. The company calls them its "most advanced models for coding and knowledge work" and says their research skills offer "an early glimpse of how AI models will contribute to scientific progress."

The headline change is cost. Anthropic says Fable 5.1 reaches similar or better results than Fable 5 at low and medium effort settings while charging noticeably less to run. For a typical workload the company puts the saving at about 25 percent, driven mainly by cheaper cached reads. For heavily agentic tasks, where a model reads and rereads large amounts of context, it rises to as much as 45 percent. The rest of the pricing is unchanged.

Where the work got better

Anthropic points to long-running tasks as the clearest area of progress. In one example it cites, Fable 5.1 traced the cause of a rare crash on the company's own systems that its engineers, and every other model it had tried, failed to pin down for years. That kind of problem, where a model has to hold a question in view across many steps rather than answer in one pass, is where recent releases have been competing hardest.

The claim is Anthropic's own, and the company has not yet published the full benchmark tables that would let outsiders check it. What is easy to check is the bill, and that is where most developers will feel the difference first.

Two models, one engine

Fable 5.1 is generally available now. Mythos 5.1 runs on the same underlying model but adds heavier safeguards around sensitive areas such as cybersecurity and life sciences, and Anthropic is limiting it to its trusted access programs rather than opening it to everyone. The split lets the company put the new capability in wide circulation while keeping tighter control over the uses it worries about most.

A faster clock

The upgrade arrives only a short time after Fable 5, part of a broader shift across the field toward frequent point releases rather than occasional big leaps. Rivals are moving on a similar cadence, and much of the contest now runs through coding and cost rather than headline scores. For the people actually building on these models, the immediate question is refreshingly concrete: does the same work cost meaningfully less to run than it did last week? Anthropic is betting the answer is yes.