Small active-parameter models are becoming the shape of local agents, argues Laurence Moroney
lmoroney · x · 2026-10-15
Laurence Moroney argues this week's releases converge on one trend: local agents are settling on MoE models with large totals but few active parameters, while phone-class AI happens at 1B and below. Speed tracks active params; memory tracks total params — so big-total/small-active models give big-model knowledge at small-model per-token cost, ideal for token-heavy agent loops. Examples: Liquid AI's LFM2.5-24B-A2B (24B total, 2B active, 64 experts top-4 routing, 131K context, claimed 184 tok/s on M5 Max, 114 on Ryzen AI Max+ 395), MiniCPM-V 1B on phones, and a faster Gemma NPU build.
Related event: Small-Activation Models Emerge as the Norm for On-Device AI Agents(2 posts)→
More from coding & agent
- HYSET scores tool sets instead of single tools, lifting agent task success to 69.9% — DeepLearningAI · 2026-10-16
- Dan Wahlin shows how to verify agent-built features in GitHub Copilot's built-in browser — DanWahlin · 2026-10-16
- Amazon would drown in money with a headless shop for agents, argues dev — djcows · 2026-10-15
- poteto open-sources pstack, a skill set for rigorous autonomous coding agents — DanielLockyer · 2026-10-15
- Geoffrey Huntley proposes the Remix Model: closed to PRs, open to remixing — kieranklaassen · 2026-10-15
- E2B sandbox usage doubled to 2B in 5 months, 61% of traffic now from agents — FinanceYF5 · 2026-10-15