On-device AI settles on small active-parameter models: phone-class agents at 1B and below
lmoroney · x · 2026-10-15
Laurence Moroney observes that local agents are converging on models with only a few active parameters, with phone-class AI now happening at 1B and below. Signals: Liquid's 2B-active model, MiniCPM-V 1B running on phones, and a faster Gemma NPU build — all pointing the same way toward small-active-parameter models as the default for on-device agents.
Related event: Small-Activation Models Emerge as the Norm for On-Device AI Agents(2 posts)→
More from Infra
- NVIDIA ships two Nemotron-3.5 open models in one day: Super for H100, Bolt for RTX — kuchaev · 2026-10-16
- Anthropic now merges 30,000 code changes a day, reviews code by 'intent' not lines — Miles_Brundage · 2026-10-16
- TSMC guides ~$176B revenue for 2026, audits customer buildouts and power before expanding — BenBajarin · 2026-10-16
- Physion Labs unveils NexTron: up to 20.9x SFT throughput vs DeepSpeed on H100 — du_yilun · 2026-10-16
- Dreamscale launches cloud inference for physical AI: 8+ robots per H100, 320ms DreamZero latency — arnie_hacker · 2026-10-16
- Google expands Android Earthquake Alerts from 98 to nearly 200 countries after 1B+ alerts — sundarpichai · 2026-10-16