<ul><li>Microsoft launched new 'open' AI models, with Phi 4 mini reasoning, Phi 4 reasoning, and Phi 4 reasoning plus being the most capable ones.</li><li>Phi 4 mini reasoning is trained on 1 million math problems for educational applications on lightweight devices, while Phi 4 reasoning is best for math, science, and coding applications.</li><li>Phi 4 reasoning plus, adapted from Phi 4 model, matches the performance levels of larger models like DeepSeek R1 and OpenAI's o3-mini on certain benchmarks.</li><li>These new models are available on Hugging Face and are designed to balance size and performance for efficient reasoning tasks on resource-limited devices.</li></ul>

Microsoft’s most capable new Phi 4 AI model rivals the performance of far larger systems

Discover more