Anthropic released Claude Fable 5.1 and Claude Mythos 5.1, including changes to agent API behavior and watermark-based text detection.
Model release Provisional 78% confidence first seen
Anthropic released two variants, Claude Fable 5.1 and Claude Mythos 5.1, described as safeguard-layer versions of the same underlying model, along with performance and pricing-related changes (including lower cache-read costs). Coverage also notes the introduction of a statistical text watermark for detecting likely model-generated or processed passages and restrictions for how “thinking blocks” are preserved for accounts created on or after Aug. 31, 2026.