Llama-3-Giraffe-70B-Instruct is a large language model released by Abacus.AI. Leveraging the PoSE and dynamic NTK interpolation training methods, it boasts an extended effective context length, capable of handling substantial volumes of textual data. Trained on approximately 1.5B tokens, the model incorporates adapter conversion techniques, applying the adapter from Llama-3-70B-Base to Llama-3-Giraffe-70B-Instruct, thus enhancing its performance.