Microsoft Research has introduced a new AI model called Orca, developed in collaboration with OpenAI. Orca is designed to imitate large language models like GPT-4, enabling it to overcome the limitations of smaller models. This 13 billion parameter-powered model can learn explanations, step-by-step thought processes, and complex instructions. It utilizes diverse imitation data to promote progressive learning and has already outperformed its predecessor, Vicuna, by 100% on zero-shot reasoning benchmarks. Orca is also 42% faster than traditional AI models on AGIEval. While it falls behind GPT-4, Orca demonstrates competitive performance on various academic exams and is expected to improve further.