Phi-4 is a state-of-the-art language model developed by Microsoft Research, designed to deliver advanced reasoning capabilities within a compact architecture. With 14 billion parameters, this dense decoder-only Transformer model is optimized for text-based inputs, particularly excelling in chat-based prompts. Trained on a diverse dataset comprising 9.8 trillion tokens—including synthetic datasets, filtered public domain content, academic literature, and Q&A datasets—Phi-4 emphasizes high-quality data to enhance its reasoning abilities. The model underwent rigorous enhancement and alignment processes, incorporating both supervised fine-tuning and direct preference optimization to ensure precise instruction adherence and robust safety measures. Released on December 12, 2024, under the MIT license, Phi-4 is tailored for applications requiring efficient performance in memory or compute-constrained environments, latency-sensitive scenarios, and tasks demanding advanced reasoning and logic. Key Features and Functionality: - Advanced Reasoning: Phi-4 is engineered to perform complex reasoning tasks, making it suitable for applications that require logical processing and decision-making. - Efficient Architecture: With 14 billion parameters, the model offers a balance between performance and resource utilization, catering to environments with memory and compute constraints. - Extensive Training Data: The model is trained on a vast dataset of 9.8 trillion tokens, including high-quality synthetic data, filtered public domain content, academic books, and Q&A datasets, ensuring a comprehensive understanding of diverse topics. - Optimized for Chat Prompts: Phi-4 excels in generating coherent and contextually relevant responses to chat-based inputs, enhancing user interaction experiences. - Safety and Alignment: The model incorporates supervised fine-tuning and direct preference optimization to adhere to instructions accurately and maintain robust safety measures. Primary Value and User Solutions: Phi-4 addresses the need for a powerful yet efficient language model capable of advanced reasoning in resource-constrained environments. Its optimized architecture and extensive training enable developers to integrate sophisticated AI capabilities into applications without compromising performance. By focusing on high-quality data and safety measures, Phi-4 ensures reliable and contextually appropriate responses, making it a valuable tool for enhancing user engagement and decision-making processes in various applications.