3 ms·

The benchmark scores in the new announcement are significantly higher than for the preview model.

• samus 2y ago
  That's good news, I was highly impressed already by what that model could do, even under heavy quantization.