Key Points
- Enhanced Model Fine-Tuning: Azure AI Foundry introduces three major updates: Reinforcement Fine-Tuning (RFT) with o4-mini, Supervised Fine-Tuning (SFT) for GPT-4.1-nano and Llama 4 Scout.
- Real-World Impact: These updates help organizations build customized AI systems, as seen in DraftWise’s 30% improvement in legal contract review quality.
- Availability: RFT with o4-mini is coming soon, while SFT for GPT-4.1-nano and Llama 4 Scout are available now on Azure AI Foundry.
Microsoft Boosts Azure AI Foundry with New Fine-Tuning Capabilities
In a significant update for Windows Azure and Microsoft’s cloud-based AI services, Azure AI Foundry is introducing three major enhancements to model fine-tuning. These updates are designed to empower organizations to create highly customized, domain-adapted AI systems that can drive real-world impact.
Reinforcement Fine-Tuning (RFT) with o4-mini: A New Era for Adaptive AI
- What’s new: RFT is coming soon to the o4-mini model, a first-of-its-kind reasoning model that can be fine-tuned.
- Why it matters: RFT enables models to learn from experience, aligning with complex business logic by rewarding desired outcomes and penalizing undesired ones. This is particularly beneficial for dynamic, high-stakes environments where adaptive reasoning and contextual awareness are crucial.
- Real-world example: DraftWise, a legal tech startup, leveraged RFT to enhance its contract review AI, achieving a 30% improvement in search result quality. This allowed lawyers to work more efficiently, demonstrating the practical application of Azure’s AI capabilities.
Supervised Fine-Tuning (SFT) for GPT-4.1-nano: Efficiency Meets Customization
- What’s new: SFT is now available for the GPT-4.1-nano model, a small but powerful foundation model ideal for cost-sensitive, high-throughput workloads.
- Why it matters: With SFT, organizations can tailor the model to their specific terminology, tone, and workflows, ensuring enterprise-grade output without compromising speed or efficiency. This model is perfect for large-scale deployments, such as customer service automation, where low latency and cost-effectiveness are key.
Llama 4 Scout Fine-Tuning: Unlocking Open-Source Potential
- What’s new: Azure AI Foundry now supports fine-tuning of Meta’s Llama 4 Scout, a cutting-edge, 17-billion parameter model boasting an industry-leading 10-million token context window.
- Why it matters: This enhancement brings the power of open-source AI to the managed compute environment of Azure, allowing for deeper customization and the ability to run on a single H100 GPU. This makes high-performance AI more accessible to a broader range of users and applications.
Getting Started with Azure AI Foundry
These updates reinforce Microsoft’s commitment to democratizing AI and driving innovation through its cloud services. Whether you’re looking to enhance decision-making with RFT, scale efficiently with SFT, or explore the possibilities of open-source AI with Llama 4 Scout, Azure AI Foundry provides the tools to build smarter, more adaptable AI systems.
- Reinforcement Fine-Tuning with o4-mini will be available soon in East US2 and Sweden Central.
- Supervised Fine-Tuning with GPT-4.1-nano and Llama 4 Scout fine-tuning are available now in North Central US and Sweden Central, and through Azure Machine Learning, respectively.
As Microsoft continues to push the boundaries of what’s possible with cloud-based AI, these enhancements pave the way for more sophisticated, customized solutions that can transform industries. Stay tuned for more updates and explore how Azure AI Foundry can help you build the future of AI, tailored to your organization’s unique needs.
Read the rest: Source Link
You might also like: Why Choose Azure Managed Applications for Your Business & How to download Azure Data Studio.
Remember to like our facebook and our twitter @WindowsMode for a chance to win a free Surface every month.