Anthropic launches Claude Fable 5.1: smarter, faster, and much cheaper

Anthropic has unveiled Claude Fable 5.1 and Claude Mythos 5.1. Three months after the launch of Fable 5, the company describes both as its most advanced models for programming and knowledge work. The combination of better performance, greater intelligence, and a lower price is, without a doubt, especially interesting.

Better coding and debugging

Fable 5.1 surpasses, of course, Fable 5 and Opus 5, but also GPT-5.6 Sol, in several key benchmarks. According to Anthropic, it’s designed for long-running tasks, where agents become much more important, and it arrives after the return of Claude Fable 5 and while the company continues to make Mythos available, the same model but with fewer safety restrictions, only for certain trusted companies.

Anthropic explains that Fable 5.1 is designed to find the root causes of errors, so we can expect better results in projects that require highly precise code over long stretches of work.

Much cheaper for the most demanding tasks

The base price stays the same, but Anthropic has decided to lower the price of cache reads. The company estimates that the total cost can drop by 25% in the most common workloads and fall by around 45% in the most demanding agentic jobs.

The update arrives with the existing safeguards, although more visible and with the invisible watermark for Claude’s text outputs.