Session: When Smaller Is Better: Using Small Language Models to Power Big Ideas
While Large Language Models (LLMs) have dominated the headlines, a quieter revolution is brewing at the edge. Small Language Models (SLMs) offer an exciting alternative- providing fast, efficient, and surprisingly capable solutions for specific use cases where LLMs may be overkill, impractical, or too resource-intensive.
In this talk, we’ll explore what SLMs are, how they differ from their larger cousins, and why developers and product teams should consider them for their applications. We’ll look at real-world use cases where SLMs excel—from on-device inference on mobile and embedded platforms, to privacy-first applications and cost-effective deployment at scale. Along the way, we’ll share practical heuristics to help you decide when an SLM is the right fit. The session will include a live demo running a language model directly on a mobile device- showcasing the versatility and potential of SLMs in the real world.
Attendees of this talk will learn about:
- What is a Small Language Model (SLM), and how it compares to an LLM
- The benefits and trade-offs of using SLMs in production environments
- Scenarios where SLMs outperform or outshine LLMs (e.g., privacy, latency, cost, edge computing)
- Practical heuristics for identifying when to reach for an SLM over an LLM
- Open source options for running SLMs on mobile devices
