AI language model

H2O-Danube-1.8B is a 1.8B language model trained on 1T markers, following the core principles of LLama 2 and Mistral. Although our model uses significantly fewer total markers when training than similarly sized reference models, it exhibits very competitive metrics on multiple benchmarks. Additionally, we release a chat model trained with supervised fine-tuning and direct preference optimization. We open source H2O-Danube-1.8B under the Apache 2.0 license to further democratize large language models and economically benefit a wider audience.