As 70,000 tech leaders, investors, and founders descend on Lisbon for Web Summit, a fundamental question hangs in the air: is the race for ever-larger AI models over? A powerful counter-movement says yes, and it’s a debate worth watching.
The Era of 'More Is More'
For the past
few years, the story of artificial intelligence has been a simple one: bigger is better. Tech giants poured billions into creating massive, general-purpose models with trillions of parameters, each new version more capable and awe-inspiring than the last. This 'frontier model' philosophy was driven by the idea that sheer scale would unlock new, unforeseen abilities. The results were impressive, producing AI that could write, code, and create with stunning fluency. But this power came at a staggering cost. Training and running these behemoths requires nation-state levels of computational power and energy, concentrating the technology in the hands of a few well-funded corporations and raising the barrier to entry for everyone else. For a long time, the industry accepted this as the only path forward. That assumption is now being challenged.
The Case for Small and Specialized
A quieter, more pragmatic revolution has been brewing. The rise of Small Language Models (SLMs) presents a compelling alternative. Ranging from one to thirteen billion parameters, these models are designed to be compact, efficient, and specialized. The core argument is that most real-world business tasks—like summarizing reports, classifying customer feedback, or powering a chatbot—don't require a model that also knows the history of the Ottoman Empire. By 2026, SLMs have become remarkably capable, delivering 90% of a frontier model's performance on specific tasks for a fraction of the cost. This shift is enabled by better training techniques and specialized hardware, like the AI-focused chips now common in smartphones. The benefits are transformative: lower costs, faster response times, enhanced privacy by running on-device, and greater control for businesses to fine-tune models on their own private data.
Lisbon's Center Stage
This very debate is set to be a major theme at this year's Web Summit in Lisbon, which runs from November 9–12. The conference agenda reveals a clear focus on this strategic pivot from brute force to intelligent efficiency. A session titled 'Beyond scale: Why the future of AI is efficient' features Dahai Li, the CEO of ModelBest, who will argue for 'high–density' models that can run on everyday devices, not just in the cloud. Another talk, 'Less is more: AI's next race is efficiency,' will see venture capitalist Rocio Wu break down where the industry can find a competitive edge with fewer resources. Even the dedicated 'AI Summit' track is structured around the return on investment for AI and the competition between open and closed models, underscoring that the conversation has moved from pure capability to practical, economic reality.
What This Means for Business
This isn't just an abstract technical argument; it’s a strategic crossroads with real–world implications. The 'bigger isn't always better' movement signals a maturing of the AI market. For businesses, it means the power of AI is becoming more accessible and affordable. Instead of being locked into expensive, one–size–fits–all APIs from a handful of providers, companies can adopt a more flexible, hybrid approach. They can use cost–effective SLMs for the bulk of their routine AI tasks while reserving the powerful, expensive frontier models for complex problems that truly require their reasoning power. This shift democratizes innovation, allowing smaller companies and startups to build competitive AI–powered products without needing massive infrastructure. It puts the focus back on finding the right tool for the job, rather than simply paying for the biggest hammer.
