Sonnet 5 is a token monster!

I started using Sonnet 5 (in Cursor and Claude Code) last night and have to say that I'm seriously impressed. It's fast and seems to be very thorough, but I've been shocked at the number of tokens it chews its way through.

Having spent quite a while carefully checking the thinking tokens, I can see that the breadth of its checking is super wide and doesn't just take the first plausible path and dive in executing. For the type of work that I'm doing (complex architecting work in a large codebase), it's definitely worth the cost as mistakes in the architecting work tend to be extremely expensive to fix when it's too late.

How are other people's experiences? As I've had time sensitive work to do, I've only been using it on max thinking level to really see the capabilities in comparison to Opus 4.8 and GPT-5.5. Curious to know if others are getting great results with a lower thinking level?