Anthropic introduced Claude Fable 5.1 and Claude Mythos 5.1 on September 1, 2026, describing them as its most advanced models yet for coding and knowledge work. The new models improve performance across multiple benchmarks while cutting costs by roughly 25-45% compared to their predecessors. Anthropic said Claude Code saw about 60% fewer cybersecurity false positives, and the release adds invisible watermarking for AI-generated text, with a detection API launching in private preview.
