AI start-ups turn to smaller models in search of cheaper scale

A growing number of AI start-ups are shipping specialised models rather than chasing the largest general-purpose systems.

The move reflects pressure from enterprise customers who want predictable costs, faster responses, and deployment options that keep sensitive data close to their own systems.

Smaller models can be easier to evaluate and tune for regulated workflows, though founders caution that the approach still requires careful monitoring and strong product judgment.

Venture investors say the next competitive advantage may come less from raw parameter counts and more from distribution, data partnerships, and operational reliability.