Microsoft Unveils ‘Phi-3-mini’, a Cost-Effective Lightweight AI Model

On Tuesday, Microsoft launched Phi-3-mini, a lightweight artificial intelligence model aimed at broadening its client base with more affordable options. This model is the first in a series of three small language models (SLMs) that Microsoft plans to release, anticipating significant global and workplace impact. SLMs are tailored to handle simpler tasks, catering especially to companies with fewer resources.

“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.

Immediately available, Phi-3-mini can be accessed through Microsoft’s Azure AI model catalog, the machine learning platform Hugging Face, and Ollama, which allows local machine model operation. Additionally, the model is compatible with Nvidia’s Inference Microservices and is optimized for use with Nvidia’s GPUs. This release follows Microsoft’s recent $1.5 billion investment in the UAE-based AI firm G42 and a partnership with French startup Mistral AI to integrate their models into Azure’s cloud platform.