It's Luna, their small, low-intelligence model. But still impressive to see.

I also found this screenshot on Reddit: The model spent 26 seconds for some reason 🤔

you are viewing a single comment's thread
view the rest of the comments
[–] 13 points 1 week ago

Success at letter counting is not a particularly good way of predicting model capabilities.

  • source