Azure AI Foundry Introduces New Fine-Tuning Models

Azure AI Foundry has announced significant enhancements to its model fine-tuning capabilities, . These updates include the introduction of Reinforcement Fine-Tuning (RFT) with the o4-mini model and Supervised Fine-Tuning (SFT) for the GPT-4.1-nano model.

Reinforcement Fine-Tuning with o4-mini is designed to improve model decision-making by aligning behavior with complex business logic. This technique rewards accurate reasoning and penalizes undesirable outputs, making it suitable for dynamic or high-stakes environments. The o4-mini model, which will soon be available, is expected to enhance adaptive reasoning and contextual awareness.

Supervised Fine-Tuning is now available for the GPT-4.1-nano model, a small yet powerful foundation model optimized for high-throughput, cost-sensitive workloads. This allows for the customization of model responses to align with company-specific tone and workflows, making it ideal for applications like customer service bots and internal knowledge assistants.

Additionally, Azure AI Foundry now supports fine-tuning of Meta's Llama 4 Scout model, a cutting-edge open-source model with a large context window and high parameter count. This model is available for fine-tuning in Azure AI Foundry and Azure Machine Learning components, offering deeper customization options.