Microsoft has launched its latest AI models, called Phi-4, designed to understand speech, images, and text at the same time. This is Microsoft's first multimodal AI model, meaning it can process different types of information together.

It has 5.6 billion parameters and helps with tasks like speech recognition, image analysis, and text processing without needing separate models.

Microsoft says the Phi-4-Multimodal model performs better than other AI models like WhisperV3 and SeamlessM4T-v2-Large in recognizing speech and translating languages. However, it is not as strong as larger models like GPT-4o when answering factual questions. Microsoft plans to improve this in future versions.

Along with the main model, Microsoft has also introduced a smaller version called Phi-4 Mini. This model has 3.8 billion parameters and is focused on text-based tasks. It is designed for reasoning, math, coding, and following instructions. It also supports up to 128,000 tokens, allowing it to handle longer pieces of text efficiently.

Another important feature of the Phi-4 Mini model is its ability to connect with external tools and APIs using function-calling features. This makes it useful for applications like smart assistants, automation, and financial data analysis.

With these new AI models, Microsoft aims to make artificial intelligence more powerful, efficient, and accessible. As AI technology continues to grow, future updates are expected to bring even better performance and features.