Microsoft unveiled Phi-3-mini, a low-resource artificial intelligence (AI) model, in an effort to appeal to a wide range of customers.
Silicon claims that this lightweight AI provides Azure users with a more affordable option.
Microsoft GenAI corporate vice president Misha Bilenko presented Phi-3 in a blog post, calling it “a family of open AI models developed by Microsoft.”
“The most capable and economical small language models (SLMs) available, outperforming models of the same size and next size up across a variety of language, reasoning, coding, and math benchmarks,” according to Bilenko, are Phi-3 models.
Although large language models (LLMs) are widely recognized as a conventional artificial intelligence (AI) solution for handling intricate tasks, their magnitude necessitates particular computational resources for their operation. As a result, Microsoft has created a range of short language models (SLMs), which are trained on smaller datasets and have many of the same features as large language models (LLMs).
Microsoft is planning to release three small language models (SLMs), the first of which is Phi-3-mini. Phi-3-mini, according to the manufacturer, measures 3.8 billion parameters and outperforms models twice its size.
In the upcoming weeks, the Phi-3 family will grow by adding more models. Furthermore, according to Bilenko, Phi-3-small and Phi-3-medium will be accessible in the Azure AI model portfolio.
Bilenko stated that “on key benchmarks, Phi-3 models significantly outperform language models of the same and larger sizes.” “Phi-3-medium and Phi-3-small outperform much larger models, including GPT-3.5T, and Phi-3-mini outperforms models twice its size.”
