Building the future with small language models

Small language models are emerging as a defining shift in how real-world AI will be used and deployed. Traditional reliance on large cloud models has created barriers in cost, data privacy, latency, and operational complexity. At the same time, new classes of desktop and on-prem hardware such as compact AI workstations and personal AI supercomputers have made it possible for capable AI systems to run locally and privately.

The session breaks down what small language models are and why they are becoming practical now, largely due to recent changes in local AI hardware. It looks at how developers and small organizations are starting to run capable models on their own machines instead of relying entirely on cloud platforms. Along the way, the talk references real tools, real hardware, and current limitations, giving attendees a grounded view of what is possible today and what directions this shift is likely to take.