New2 developments · 1 source · over 6 daysLiquid AI opens d1 decision models for edge inferenceHow it developedOctober 7, 2026Hugging Face BlogWhere it startedLiquid AI opens d1 decision models for edge inferenceOctober 13, 2026Hugging Face BlogLiquid AI releases efficient on-device agent modelLiquid AI releases LFM2.5-24B-A2B, an efficient on-device agent model with 131K context that decodes at 184 tok/s on Apple M5 Max.