Microsoft has introduced new AI models in its Phi family, emphasizing reasoning capabilities that allow for better fact-checking and problem-solving. The Phi 4 mini reasoning is optimized for educational uses and trained on a million synthetic math problems. The Phi 4 reasoning model, with 14 billion parameters, leverages high-quality web data for applications in math and science, while Phi 4 reasoning plus aims for improved accuracy, closely matching the performance of larger models. All models support developers on the Hugging Face platform, promoting wider accessibility and application.
The Phi 4 family of models represents an evolution in Microsoft's AI offerings, focusing on reasoning capabilities that enhance problem-solving across various complex tasks.
These new models are designed not just for size, but for performance, achieving better accuracy and versatility in educational and technical applications.
Collection
[
|
...
]