Introduction to Microsoft’s New AI Models

Microsoft has introduced several new “open” AI models as of Wednesday, with the most capable model being competitive with OpenAI’s o3-mini on at least one benchmark.

Details of the New Models

All the newly released permissively licensed models, including Phi 4 mini reasoning, Phi 4 reasoning, and Phi 4 reasoning plus, are categorized as “reasoning” models. This means they are capable of spending more time fact-checking solutions to complex problems. These models expand Microsoft’s Phi “small model” family, which was launched a year ago to provide a foundation for AI developers building applications at the edge.

Phi 4 Mini Reasoning Model

The Phi 4 mini reasoning model was trained on approximately 1 million synthetic math problems generated by Chinese AI startup DeepSeek’s R1 reasoning model. With around 3.8 billion parameters in size, Phi 4 mini reasoning is designed for educational applications, such as “embedded tutoring” on lightweight devices, according to Microsoft.

Understanding Model Parameters

Model parameters roughly correspond to a model’s problem-solving skills. Generally, models with more parameters perform better than those with fewer parameters.

Phi 4 Reasoning Model

Phi 4 reasoning, a 14-billion-parameter model, was trained using “high-quality” web data as well as “curated demonstrations” from OpenAI’s aforementioned o3-mini. It is best suited for math, science, and coding applications, according to Microsoft.

Phi 4 Reasoning Plus Model

As for Phi 4 reasoning plus, it is Microsoft’s previously released Phi-4 model adapted into a reasoning model to achieve better accuracy on particular tasks. Microsoft claims that Phi 4 reasoning plus approaches the performance levels of R1, a model with significantly more parameters (671 billion). The company’s internal benchmarking also shows Phi 4 reasoning plus matching o3-mini on OmniMath, a math skills test.

Availability of the Models

Phi 4 mini reasoning, Phi 4 reasoning, and Phi 4 reasoning plus are available on the AI dev platform Hugging Face, accompanied by detailed technical reports.

Conclusion and Future Prospects

“Using distillation, reinforcement learning, and high-quality data, these [new] models balance size and performance,” wrote Microsoft in a blog post. “They are small enough for low-latency environments yet maintain strong reasoning capabilities that rival much bigger models. This blend allows even resource-limited devices to perform complex reasoning tasks efficiently.”