Microsoft on Tuesday shared Phi-3, an update to its small language model that it says is capable of handling many tasks that had been thought to require far larger models.   "Phi-3 is not slightly cheaper; it's dramatically cheaper; we're talking about a 10x cost difference compared to the other models out there with similar capabilities," said Sébastien Bubeck, Microsoft's vice president of GenAI research.

Free access to multiple LLMs at Merlin

Get Merlin

SLMs are designed to perform simpler tasks, making them easier to use by companies with limited resources, the company said.

Phi-3-mini will be available immediately on Microsoft cloud service platform Azure's AI model catalog, machine learning model platform Hugging Face, and Ollama, a framework for running models on a local machine, the company said.

Jaspreet Bindra, Founder, of TechWhisperer UK Limited, commented, "The newest such ‘Small Language Model’ comes from Microsoft with Phi-3 Mini. It is a lightweight artificial intelligence model developed as part of its small language model (SLM) family. It is the first of three SLMs that Microsoft plans to launch shortly, with the other two being Phi-3 Small and Phi-3 Medium. The Phi-3 Mini has a capacity of 3.8 billion parameters and is designed to perform simpler tasks, making it more accessible and affordable for businesses with limited resources.

Free access to multiple LLMs at Merlin

Get Merlin

Phi-3 Mini has been trained on a dataset that is smaller than that of large language models such as GPT-4. It is a part of Microsoft's broader initiative to introduce a series of SLMs that are tailored for simpler tasks, making them ideal for businesses with fewer resources. This approach promises to reduce costs, make the models faster as they are on the edge, and promise more enterprise and consumer use cases of generative AI."   Phi-3 empowers developers to build generative AI applications at the edge—more efficiently, with better latency and lower costs. Microsoft has been developing a series of SLMs that offer many of the same capabilities found in LLMs but are smaller in size and are trained on smaller amounts of data. “We started this journey a year ago trying to build small language models, trying to see, how far we can go with maybe just 1 billion parameters,” said Sebastien Bubeck, VP at Microsoft GenAI.

Microsoft’s competitors have their own small AI models as well, most of which target simpler tasks like document summarization or coding assistance. Google’s Gemma 2B and 7B are good for simple chatbots and language-related work.