Anthropic launches Claude Fable 5.1 and Mythos 5.1 with big coding gains

Anthropic, a US AI company, released two new models on September 1: Claude Fable 5.1 and Claude Mythos 5.1. They share the same base but differ in safety: Fable 5.1 is open to everyone, while Mythos 5.1 is restricted to approved cybersecurity and life sciences groups.

In coding tests, Fable 5.1 scored 55.8% on Terminal-Bench 4.0, and Mythos 5.1 hit 60.9% with looser safety, both beating previous models and OpenAI's GPT-5.6 Sol. On the science agent test, Fable 5.1 got 52.6%, up from 24.7% for the old version.

Pricing: cache reads dropped 75% to $0.25 per million tokens, and typical workloads cost about 25% less than the older model, up to 45% less for heavy agent use. Base prices stay the same: $10 per million input tokens, $50 per million output.

Anthropic also introduced enterprise frontier safeguards with a zero-data-retention option, rolling out this fall. In science, Mythos 5.1 showed big wins in protein design, Venus mapping, and GPU speedups.