Anthropic Launches Fable 5.1 and Mythos 5.1 with 75% Cache Read Price Cut

My take: A 75% cut on cache read pricing is not a minor price adjustment: for teams building agentic applications on Anthropic's API, it can mean up to 45% less on the monthly bill. That was a core part of yesterday's launch of Fable 5.1 and Mythos 5.1, the company's most capable model pair.

The benchmarks are also notable: Terminal-Bench-Science went from 24.7% to 52.6% compared to Fable 5, and AutomationBench from 17.1% to 31.4%, more than doubling in both cases. That said, these numbers deserve perspective: Anthropic measures its own models against its own metrics, so, as both judge and party, waiting for independent validation before treating these numbers as definitive is the right call.

What does not require external validation is the API cost cut: it applies now. Do you know exactly what your current agentic stack would cost with the new pricing, or is this the moment to run that calculation?