Small-Activation Models Emerge as the Norm for On-Device AI Agents

Former Google AI advocate Laurence Moroney argues that on-device agents are converging on MoE models with large total but small activated parameters, with phone-level AI now achievable at 1B or fewer activated parameters.

2026-10-15 ~ 2026-10-15 · 2 related posts

• On-device AI settles on small active-parameter models: phone-class agents at 1B and below — lmoroney · 2026-10-15
• Small active-parameter models are becoming the shape of local agents, argues Laurence Moroney — lmoroney · 2026-10-15