Microsoft Foundry supports Reinforcement Fine-Tuning (RFT) with the o4-mini model in Azure OpenAI, enabling adaptive learning for context-aware applications. Alongside RFT, we introduced Global Training, allowing fine-tuning from 12+ Azure regions at reduced costs, and promoted Global Standard and Regional Provisioned Throughput deployment types to general availability, offering more cost and data residency flexibility.
To support rapid experimentation, a new Developer Tier was launched, providing free 24-hour hosting for fine-tuned GPT-4.1 and GPT-4.1-mini models with pay-per-token pricing. This tier is ideal for quick testing and model comparison before production.
Additionally, we extended support for fine-tuned GPT-4o and GPT-4o-mini models, allowing continued deployment for 12 months post-retirement at no extra cost, providing continuity for long-term projects.
These updates reflect Microsoft’s commitment to making fine-tuning more accessible, scalable, and production-ready for custom AI solutions.