Microsoft unveiled Phi-4 artificial intelligence capable of solving complex mathematical problems

Microsoft has introduced a new member to its family of Phi AI models. The software giant says that this model named Phi-4 In several fields and in particular Solve math problems It has better performance than the previous models of this series.

The Phi-4 AI has been made available in a very limited manner since last night. According to a TechCrunch report, it can actually only be used with the company’s Azure AI Foundry development platform and only for research purposes licensed by Microsoft.

Microsoft’s Phi-4 Artificial Intelligence Performance

Phi-4 is Microsoft’s newest small language model, which is 14 billion parameters in size, and will compete with other small models including GPT-4o Mini, Gemini 2.0 Flash, and Claude 3.5 Haiku. These AI models are often faster and cheaper to implement, but the performance of small language models has gradually increased over the past few years.

Microsoft attributes the improved performance of its Phi-4 AI model to the use of a “high-quality synthetic dataset” alongside a high-quality dataset of human-generated content and some unspecified post-training improvements.

Another point that TechCrunch mentioned at the end of its report is that Phi-4 is the first model of this Microsoft family that was released after the departure of Sebastian Babak from the technology giant. Babak, formerly vice president of artificial intelligence at Microsoft and a key figure in the development of Phi models, left Microsoft in October to join OpenAI.