Training for the new V2 model in progress. 10% of the training already exceeds the previous Finnish-NLP/llama-3b-finnish and Finnish-NLP/llama-7b-finnish models in the FIN-bench evaluation.