Distilled version of Llama 3.1 8B using reasoning data generated by DeepSeek R1 for enhanced performance.