Llama-3.1-Tulu-3-70B-DPO is part of the Tülu3 model family, offering comprehensive guidelines for modern post-training techniques. This model family aims to achieve state-of-the-art performance across various tasks beyond chat, such as MATH, GSM8K, and IFEval. It is trained on publicly available, synthetic, and human-created datasets, mainly in English, and complies with the Llama 3.1 community licensing agreement.