Anthropic Claude Fable 5.1, Claude Mythos 5.1 Models Announced for Advanced Coding, Scientific Research

Anthropic has introduced Claude Fable 5.1 and Claude Mythos 5.1 as its latest models designed for advanced coding and knowledge work. Fable 5.1 is confirmed to be generally available, while Mythos 5.1 is accessible through trusted programmes. The two models work on the same system, but have different safety safeguards. As per Anthropic's benchmarks, Claude Fable 5.1 surpassed Fable 5 and Opus 5 in science and agentic coding tasks by scoring 52.6 percent on Terminal-Bench-Science and 73.4 percent on CursorBench.

Anthropic Claude Fable 5.1 and Mythos 5.1

In a blog post, Anthropic announced the Claude Fable 5.1 and Claude Mythos 5.1 models for coding, knowledge work and long-running problem-solving tasks. As mentioned, Claude Fable 5.1 is generally available, while Mythos 5.1 is available only through Anthropic's trusted access programs. The company says its safeguards are specifically designed to support work in cybersecurity and the life sciences. It adds that the new model is designed based on user feedback on price, data retention, and safeguards.

The Claude Mythos 5.1 is identical to Fable 5.1, but it has more permissive safeguards for vetted users and organisations working with cybersecurity and life sciences restrictions. It will be offered through Anthropic's Cyber Verification Program and Life Sciences Verification Program.

Fable 5.1 is confirmed to cost an estimated 25 percent less than Fable 5 for typical workloads due to reduced cache read pricing, wherever usage is billed by token. For highly agentic work, it is said to cost 45 percent less than Fable 5.

Anthropic's benchmarks show that Fable 5.1 outperforms Fable 5 and Opus 5 in science and agentic coding tasks. It scores 52.6 percent on Terminal-Bench-Science. Fable 5 scored 24.7 per cent, and Opus 5 scored 29 percent. Fable 5.1 received 73.4 percent on CursorBench compared with 70.5 percent for Fable 5 and 70 percent for Opus 5.

Anthropic said Fable 5.1 has been used in complex real-world problems. It claims that in a test by investment firm Millennium, Fable 5.1 identified a long-standing software crash caused by a bug in an external vendor library. It also claimed to have tested its capabilities in Computational analysis and modelling, Scientific research, and computational biology. Mythos 5.1 said to have achieved high-affinity protein design with hit rates near 50 percent. It claimed to have accelerated seven open-source deep learning models by up to 2.5 times with identical outputs.