Anthropic unveils Claude Fable 5.1 that can work for days on problems humans couldn't solve

Anthropic has unveiled Claude Fable 5.1, a new version of its flagship AI model that the company says can handle complex coding, research and other tasks for much longer without human intervention.

The company also introduced Claude Mythos 5.1, a more restricted version aimed at advanced cybersecurity and scientific research.

Anthropic says Fable 5.1 is particularly better at tasks that require an AI system to work through several steps rather than simply answer a question and stop.

One of the biggest improvements comes in scientific work. On Anthropic's Terminal-Bench-Science test, Fable 5.1 scored 52.6%, more than twice the 24.7% recorded by its predecessor, Fable 5.

Anthropic said the model has also been used on problems that required long-running investigations. In one example, the company said Fable 5.1 helped identify the cause of a software crash at Millennium that engineers had struggled to explain for four to five years.

The new model is not more expensive on standard token pricing, with Anthropic keeping the price at $10 per million input tokens and $50 per million output tokens. However, the company says typical workloads can be around 25% cheaper, while highly agentic tasks can cost up to 45% less.

There is also a significant change in how the model handles cached information. Anthropic has cut the price of cache reads by 75%, from $1 to 25 cents per million tokens.

Mythos 5.1 will remain limited in availability, while Fable 5.1 is being offered as the broadly available model.

Anthropic is pitching the release as a step toward AI systems that can do more than generate quick answers instead working for hours or longer on difficult problems with considerably less human involvement.