Anthropic released Fable 5.1 — a model for programming and scientific research

Anthropic released Claude Fable 5.1 and Claude Mythos 5.1 — new models for programming and knowledge work. Both are built on the same foundation but differ in their safety mechanisms. Fable 5.1 is available to all users, while Mythos 5.1 is accessible only through trusted access programs for cybersecurity and biology specialists.

The company has reduced the cost of using Fable 5.1. The price for reading the cache — reprocessed context — has been reduced by 75% and is now $0.25 per million tokens. As a result, typical tasks via API have become approximately 25% cheaper, and complex agentic tasks up to 45%. Base rates remain unchanged: $10 per million input tokens and $50 per million output tokens.

According to Anthropic, in the Terminal-Bench 4.0 test, Fable 5.1 scored 55.8% compared to 42% for Fable 5. In the Terminal-Bench-Science 0.1 test, the result increased to 52.6% compared to 24.7%. In OSWorld 2.0 under strict evaluation mode — 41.7% compared to 36.1%. In Humanity's Last Exam without using tools — 60.9% compared to 57.8%.

Anthropic also updated its safety mechanism system. In the field of cybersecurity, the number of false positives has been reduced by 60% — the model can now be used to find vulnerabilities in code, but not to develop exploits. For working on biology-related tasks, the company launched a separate access program developed in collaboration with the US government.

Claude Mythos 5.1 is available through two programs — the Cyber Verification Program for cybersecurity specialists and the Life Sciences Verification Program for professionals in the life sciences.

Claude Fable 5.1 is already available on all platforms, including Amazon Web Services, Google Cloud, and Microsoft Azure. Developers can use the model via the Claude API under the name claude-fable-5-1.

Source: Anthropic