Anthropic has announced the release of its new AI models, "Claude Fable 5.1" and "Claude Mythos 5.1." According to the company, Fable 5.1 can reduce inference costs by up to 45% compared to the previous Claude 5.
Regarding the model's pricing structure, rates will be set from $0.25 to $0.40 per 1 million tokens. This represents a cost reduction of 25% to 50% compared to the conventional Claude 5.
Improvements have also been observed in coding performance. In the Terminal-Bench 4.0 benchmark, Fable 5.1 recorded a score of 55.8%, surpassing the previous Claude 5 (42.0%) and Opus 5 (52.3%). It also demonstrated a result of 73.4% in CursorBench 3.2.
On the security front, Mythos 5.1 supports AI vulnerability detection. The model is equipped with capabilities to identify vulnerabilities contained within program code.
Fable 5.1 is available through the Claude API via Amazon Web Services, Google Cloud, and Microsoft Azure platforms.
Source: "Up to 45% cost reduction" Anthropic announces "Claude Fable 5.1" with vulnerability detection support (ITmedia AI+, 2026-09-05)