Microsoft Azure AI Foundry already supports model fine-tuning. Today, Microsoft announced major improvements to model fine-tuning on AI Foundry with support for Reinforcement Fine-Tuning (RFT) and more. RFT is a new technique that uses chain-of-thought reasoning and task-specific grading to improve model performance for specific domains.
OpenAI first announced the alpha program for RFT last December. According to their early testers, RFT delivered a 40% increase in model performance when compared to standard out-of-the-box models. Microsoft today announced that RFT for OpenAI's o4-mini model is coming soon to Azure AI Foundry. The company recommends RFT in the following scenarios:
- Custom Rule Implementation: RFT thrives in environments where decision logic is highly specific to your organization and cannot be easily captured through static prompts or traditional training data. It enables models to learn flexible, evolving rules that reflect real-world complexity.
- Domain-Specific Operational Standards: Ideal for scenarios where internal procedures diverge from industry norms—and where success depends on adhering to those bespoke standards. RFT can effectively encode procedural variations, such as extended timelines or modified compliance thresholds, into the model’s behavior.
- High Decision-Making Complexity: RFT excels in domains with layered logic and variable-rich decision trees. When outcomes depend on navigating numerous subcases or dynamically weighing multiple inputs, RFT helps models generalize across complexity and deliver more consistent, accurate decisions.
Microsoft today also announced support for Supervised Fine-Tuning (SFT) for OpenAI's latest GPT-4.1-nano model, which is suitable for cost-sensitive AI applications. Fine-tuning for GPT-4.1 will be available in the coming days.
Finally, Microsoft announced support for fine-tuning Meta’s latest Llama 4 Scout 17 billion parameter model that comes with support for a 10M token context window. Also, the Llama 4 fine-tuning is available as part of Azure's managed compute offering. The fine-tuned Llama model is available in both Azure AI Foundry and as Azure Machine Learning components.
1 Comment - Add comment