Microsoft Unveils New AI Model Phi-4

Microsoft has unveiled a new generative AI model, Phi-4i, which has improvements in several areas, including solving math problems.

Microsoft Phi-4 performs better than its predecessor in a number of ways, including in solving mathematical problems, the developer claims — partly due to higher-quality training data. Phi-4 is currently available in very limited access on the Azure AI Foundry development platform — according to the license agreement, it can only be used for research purposes. The next-generation small language model has 14 billion parameters — it is positioned as a competitor to projects such as GPT-4o mini, Gemini 2.0 Flash, and Claude 3.5 Haiku. These small AI models are faster, cheaper to maintain, and their quality has recently skyrocketed.

Microsoft attributes Phi-4’s success to the fact that it was trained using high-quality human-generated content alongside “high-quality synthetic datasets” and that some model refinement was done after training.

Phi-4 is currently only available on Microsoft’s recently launched Azure AI Foundry development platform and for research purposes only.