Anthropic released Claude Mythos 5.1 on Monday, a restricted model priced at ten dollars per million input tokens and fifty dollars per million output tokens that only vetted organisations in the United States can use. The launch matters because it creates a two-tier market: general subscribers on Free, Pro, Max or Team plans cannot select the model at any price, while approved cyber defenders and life-science researchers get a version of the same underlying system that powers Claude Fable 5.1 with looser safeguards for their specialist work.
The pricing ladder has compressed fast
The current rates represent a sharp drop from the April preview, when Anthropic quoted twenty-five dollars for input and one hundred twenty-five for output after a hundred-million-dollar credit programme expired. Mythos 5 arrived in June at the same ten-and-fifty levels that 5.1 now carries, but cache reads cost a dollar per million tokens then. The new version cuts that to twenty-five cents, a seventy-five percent reduction, while cache writes hold at twelve dollars fifty for five minutes and twenty dollars for one hour. Batch jobs still run at half the standard rates.
Access stays gated and geography-locked
Anthropic grants entry through trusted programmes that require separate API or cloud-platform billing; approval does not waive usage fees. The company restored access for a group of US organisations on July 1 after export-control questions had disrupted the June rollout, and the September release keeps the same geographic limit. About a hundred fifty organisations across more than fifteen countries participated in the expanded June phase, but the current production version is described as limited to selected US entities.
Same model, different guardrails
Mythos 5.1 and Fable 5.1 share the same base weights, a one-million-token context window and a 128,000-token output ceiling. Both accept text and images and return text, and both keep adaptive thinking enabled with an effort knob for reasoning depth. The distinction is purely policy: Mythos relaxes controls for approved defensive cybersecurity and biology tasks while retaining Anthropic's other safety rules. The company says the model can analyse source code, research files, screenshots, tables and document pages, though the full feature list remains partially documented.