Microsoft has released its smallest AI model, Phi-3 Mini. It is the first of three planned. This was reported by The Verge.

The AI model is already available on Azure, Hugging Face, and Ollama. It has 3.8 billion parameters and is trained on a smaller data set compared to large models like GPT-4.

Microsoft also says that Phi-3 performs better than Phi-2 and can give the same answers as ten times larger models.

In addition, the company plans to release two more models – Phi-3 Small with 7 billion parameters and Phi-3 Medium with 14 billion parameters. Parameters indicate how many complex instructions a model can understand.

Eric Boyd, Corporate Vice President of Microsoft Azure AI Platform, said that Phi-3 Mini is as powerful as large language models like GPT-3.5, but in a smaller form factor. In addition, their compact size allows them to work on personal devices such as a smartphone or laptop.

Boyd also said that when developing the Phi-3, they relied on what the previous models could do. With the Phi-1, they focused on coding, with the Phi-2 on reasoning, and with the Phi-3, they combined and improved on these.

Despite this, such compact models still cannot match the breadth of GPT-4 or other models that learn from the entire Internet.