Anthropic drops Fable 5.1 and Mythos 5.1, cutting agent costs up to 45%

Anthropic updated its model suite with Claude Fable 5.1 and Mythos 5.1, cutting agentic workload costs by up to 45%.

Anthropic released Fable 5.1 for general availability and Mythos 5.1 for trusted partners in biology and cybersecurity. Both use the same base model, but Mythos adds specialized safeguards for sensitive research. On typical workloads, Fable 5.1 cuts token costs by 25%. For multi-step agentic tasks, cheaper input cache reads push those savings to around 45%.

Why it matters: Long-running AI loops get expensive fast. Fable 5.1 delivers better reasoning while burning less budget. On the Terminal-Bench-Science benchmark, Fable 5.1 scored 52.6% accuracy, up from Fable 5's 24.7%. Investment firm Millennium even reported that Fable 5.1 pinpointed the root cause of an internal system crash that had stumped its engineers for years.

Fable 5.1 is live today. It defaults to High effort in Claude Code and Medium in Claude Cowork and on Claude.ai. If you require zero data retention, eligible accounts can enable it now ahead of Anthropic's Enterprise Frontier Safeguards rollout this fall.

AI models that solve harder problems while cutting your compute bill are a rare double win.