Hacker News Reader

The extra thinking time I assume allows it to be the best model. I saw it think for 2 minutes on a single question.

Not only extra time, but lots more internal prompts and tokens, thus potentially increasing your bill.

I'm thrilled with the results but expect 10-100x more token consumption.