Previous Granite models were on the level of first llama in my benchmarks.
I’m expecting this version to be roughly comparable to llama 2