Anthropic’s Claude Fable 5.2 and Mythos 5.0 Impress but Carry Steep Costs

The real story here is the economics of frontier capability, not the benchmarks. Models that can design physical objects blur the line between software tooling and engineering labor, which matters for teams weighing AI against hiring or traditional CAD and prototyping workflows. The open question is whether cost per useful output falls fast enough to make these models routine rather than occasional tools — early adopters wincing at bills suggests capability is currently outrunning affordability, and that gap will shape adoption more than raw performance.