Microsoft has presented a new generative model of artificial intelligence called Phi-4. It contains 14 billion parameters and demonstrates better efficiency in solving mathematical problems compared to previous versions. This was reported by TechCrunch.
Phi-4 is available on a limited basis only on the recently launched Microsoft Azure AI Foundry development platform, and for research purposes only under the Microsoft license agreement.
The Phi-4 will compete with models such as the GPT-4o mini, Gemini 2.0 Flash, and Claude 3.5 Haiku. These versions of AI models are often faster and cheaper, and their performance has been gradually increasing over the past few years.
Microsoft attributes Phi-4's performance boost to the use of "high-quality synthetic datasets" paired with classic human-generated datasets.
By the way, Phi-4 is the first model of the Phi- series that was launched after the departure of Microsoft's Vice President of AI Sebastien Boubek. He was a key figure at Microsoft in the AI department, but left the company to join OpenAI.