Tech giant Microsoft is staking a major claim in the generative artificial intelligence space, launching a lightweight and affordable AI model dubbed “Phi-3-mini” as part of a new family of smaller language models (SLMs). The unveiling of Phi-3-mini marks Microsoft’s first foray into compact, cost-optimized AI models as it looks to attract a wider range of customers and companies with limited resources to its generative AI offerings.
Unlike the large and computationally intensive foundation models like GPT-3 that have captured headlines, SLMs like Phi-3-mini are designed to handle simpler tasks and use cases in a more streamlined and economical fashion. According to Sébastien Bubeck, Microsoft’s vice president of generative AI research, Phi-3-mini delivers dramatically lower operating costs compared to conventional AI models with similar capabilities.
“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,” Bubeck stated, highlighting the potential for major savings realized through Microsoft’s optimized model architecture.
In the fierce battle to secure market share in generative AI, a technology expected to fundamentally transform how people work and live, Microsoft is betting that SLMs like Phi-3-mini will open the door to a broader swath of cost-conscious enterprises, developers, and organizations. These leaner models could prove pivotal for driving adoption beyond the major tech giants able to shoulder the substantial computing costs of training and running large language models.
Microsoft is making Phi-3-mini immediately available across multiple platforms and channels frequented by machine learning developers and AI practitioners. This includes its own Azure cloud AI model catalog, the popular open-source Hugging Face model hub, and Ollama, a framework for running models locally.
The lightweight model release represents the latest maneuver in Microsoft’s generative AI blitz, following a $1 billion investment in UAE artificial intelligence firm G42 last week. Microsoft had also previously partnered with French AI startup Mistral AI to integrate its language models into the Azure cloud platform’s offerings.
As competition in generative AI intensifies between Microsoft, Google, Amazon and a host of AI startups, the ability to deliver powerful yet affordable models could prove decisive in winning over enterprises still dipping their toes into the transformative technology. Microsoft’s strategic SLM push signals its intent to democratize AI accessibility and establish itself as the partner of choice for companies across sectors seeking on-ramps to generative AI capabilities.
