Microsoft Phi-3 AI technology Top Builders

Explore the top contributors showcasing the highest number of Microsoft Phi-3 AI technology app submissions within our community.

Phi-3 Model Family

The Phi-3 model family, developed by Microsoft, encompasses a range of small language models (SLMs) designed to offer high-quality AI capabilities with a focus on efficiency and accessibility. These models are particularly suited for applications where computational resources are limited, such as mobile devices or edge deployments. The Phi models balance performance and size, making them ideal for a variety of use cases, from natural language understanding to coding tasks.

General
Relese date	June 2023
Author	Microsoft
Website	Phi-3 open models
Type	Small Language Models

Key Models and Features

Phi-3 Mini: A compact model with 3.8 billion parameters, trained on 3.3 trillion tokens. It provides strong performance across various benchmarks, including MMLU and GSM-8K, and is capable of running locally on smartphones and other edge devices.
Phi-3 Small: Featuring 7 billion parameters, this model includes additional capabilities for handling longer context lengths (up to 128K tokens). It offers enhanced performance in reasoning tasks and is fine-tuned with supervised and preference optimization techniques.
Phi-3 Medium: A larger variant with 14 billion parameters, designed for more complex applications that require robust reasoning and data analysis capabilities.

Training and Data

The training data for Phi models is meticulously curated, combining publicly available high-quality documents, synthetic “textbook-like” data, and chat-format supervised data. This approach ensures the models have a strong foundation in reasoning, coding, and general knowledge while maintaining efficiency in processing and storage requirements.

Applications and Use Cases

Edge and Mobile Deployment: The small size and efficient design of the Phi models make them suitable for deployment on devices with limited computational power, such as smartphones or IoT devices. They can operate offline, which is crucial for applications in remote or disconnected environments.
High-Risk Scenarios: While the models are designed to minimize biases and handle sensitive data responsibly, they are not recommended for high-stakes applications like legal advice or financial decision-making without additional safeguards.

Availability and Licensing

The Phi models are available on platforms like Hugging Face and Microsoft Azure AI Model Catalog. They are released under open licenses, allowing developers to integrate them into various applications while adhering to responsible AI practices.

👉 For more detailed information, you can refer to the technical report and resources available on Microsoft Research and Hugging Face.

Edit on Github

Microsoft Phi-3 AI technology Hackathon projects

Discover innovative solutions crafted with Microsoft Phi-3 AI technology, developed by our community members during our engaging hackathons.

EdgeWise-Offline AI Content Moderation

EdgeWise is an AI-driven content moderation solution designed specifically for educational platforms. Leveraging advanced AI models, our tool operates entirely offline on edge devices such as IoT devices, ensuring robust performance even in environments with limited or no internet access. The architecture of EdgeWise consists of several key components: Synthetic Data Generation: We use OpenAI's Meta-Llama-3.2-80B-Instruct-Turbo model to generate synthetic training data tailored to specific content moderation categories such as spam, inappropriate content, and misleading information. Fine-Tuned Model: The generated data is used to fine-tune the Phi model. This fine-tuned model is lightweight, optimized for edge devices, and includes specialized LoRA adapters for efficient inference. Edge Deployment: The fine-tuned model is deployed locally on devices using a Streamlit-based application. This application is designed to work entirely offline, providing real-time text categorization and content filtering without relying on external APIs or cloud services. Privacy and Security: By processing all data locally, EdgeWise ensures that user information remains private and secure. The architecture is robust, cost-effective, and highly customizable, allowing it to adapt to various educational environments and needs. This combination of advanced AI, local deployment, and a focus on privacy makes EdgeWise an ideal solution for creating safe, secure, and inclusive online learning environments globally.

DefendX

DefendX is an AI-powered cybersecurity solution focused on safeguarding edge devices by detecting and mitigating security threats like unusual login activities, background-running malware, and abnormal traffic at routers. Leveraging cutting-edge technologies such as Llama, Phi3, and Streamlit, DefendX offers real-time anomaly detection and threat analysis tailored for edge environments. Llama's robust natural language processing and Phi3's advanced machine learning algorithms enable the model to efficiently recognize and respond to potential security breaches. Integrated with a user-friendly Streamlit interface, DefendX provides seamless interaction, allowing users to easily monitor and address threats while ensuring optimal device performance and data privacy.

SafeEdge- online education inclusive and enjoyable

SafeEdge is an AI-driven content moderation solution designed specifically for educational platforms. Leveraging advanced AI models, our tool operates entirely offline on edge devices such as tablets, laptops, and IoT devices, ensuring robust performance even in environments with limited or no internet access. The architecture of SafeEdge consists of several key components: Synthetic Data Generation: We use OpenAI's Meta-Llama-3.1-70B-Instruct-Turbo model to generate synthetic training data tailored to specific content moderation categories such as spam, inappropriate content, and misleading information. Fine-Tuned Model: The generated data is used to fine-tune the Phi-3-mini-4k-instruct model. This fine-tuned model is lightweight, optimized for edge devices, and includes specialized LoRA adapters for efficient inference. Edge Deployment: The fine-tuned model is deployed locally on devices using a Streamlit-based application. This application is designed to work entirely offline, providing real-time text categorization and content filtering without relying on external APIs or cloud services. Privacy and Security: By processing all data locally, SafeEdge ensures that user information remains private and secure. The architecture is robust, cost-effective, and highly customizable, allowing it to adapt to various educational environments and needs. This combination of advanced AI, local deployment, and a focus on privacy makes SafeEdge an ideal solution for creating safe, secure, and inclusive online learning environments globally.

AI Power Cybersecurity 4 Next Gen EndPoint Detect

Team Qerberos is committed to revolutionizing the cybersecurity landscape by leveraging cutting-edge LLMs and other AI algorithms to address complex challenges in the field. Our strategy centers on creating solutions that not only advance security protocols but also integrate seamlessly into existing infrastructures. For this hackathon, Team Qerberos have developed prototype LLM models designed for synthetic data creations and deployment across various endpoints—including smartphones, laptops, servers, and IoT devices. This model operates without disrupting current production processes, providing robust, real-time detection of malicious activities with exceptional accuracy. Team Qerberos’ approach ensures that organizations can enhance their security posture while maintaining operational efficiency.