Neither Sonnet nor Opus could solve it or get close in a minimal test I did just now, using the same prompt as above.

Sonnet: https://pastebin.com/24QG3JkN

Opus: https://pastebin.com/PJM99pdy

I think this new model is a generational leap above Claude for tasks that require complex reasoning.

Way worse than Claude for solving a cipher. Not even 1/10th as good. Just one data point, ymmv.