I used text. I tried voice and it was super dumb and made a lot of mistakes. The text mode made only one microscopic math mistake.

As for price, you can use Gemini 2.5 Pro for free in Google AI Studio. The entire 1hr+ I spent in the adventure was entirely one session and completely under free usage. I could probably done a few more hours or more without hitting daily limits.

Gemini is one of the better models right now and has an amazing 1M token context window. I’m only at 67K out of 1M context for my entire adventure.

I can compress it when I get closer.

Context window is what makes using it locally virtually impossible as 16k-64k is about the best you can hope for on consumer hardware but more importantly much smaller models.