maybe try again - this is still in beta and isn't perfect. however - its much better than the base model, llama 3.1 70b.