2 ms·

I guess we are going to be using multiple small specialized models with a reasoning model and tooling.

• selim-now 1y ago
  isn't that the premise of the Nvidia paper? https://arxiv.org/pdf/2506.02153 https://arxiv.org/pdf/2506.02153