Anthropic Launches Claude 5.1 Models, Cuts Cache Costs

Anthropic is expanding its Claude lineup as developers increasingly weigh inference costs alongside coding and reasoning performance. Claude Fable 5.1 targets general-purpose coding and complex reasoning, while Claude Mythos 5.1 is tailored to scientific research. The split underscores the company’s push toward specialized models for demanding enterprise and agentic workloads, where repeated access to long context can make prompt caching a significant part of operating expenses.

As of Sept. 2, 2026, Anthropic has released Claude Fable 5.1 and Claude Mythos 5.1, cutting cache-read prices by 75%. The company says the reduction can lower total costs for highly agentic tasks by as much as 45%. The models also improve coding and reasoning performance, reduce false refusals caused by overly aggressive safety filters, and introduce anti-distillation safeguards designed to meet regulatory requirements and deter competitors from extracting model capabilities.