What actually shipped in agent engineering, pulled from the labs, arXiv and Hacker News.
See who we followLiquid AI and Qualcomm Technologies are integrating personal AI context capabilities into Snapdragon-powered devices.
Liquid AI introduced Pipette, a benchmarking suite for evaluating on-device intelligence systems.
LFM2.5-DSpark achieves up to 3.2x faster inference from H100 to MacBook.
LFM2.5 Q4_0 applies quantization-aware distillation for edge device deployment.
Liquid AI released LFM2.5-VL-3B, a 3 billion parameter vision-language model optimized for edge device deployment.
MacPaw partners with Liquid AI to deploy on-device AI models for Mac users.
LFM2.5-2.6B is a model for deploying agents across diverse environments.
Liquid AI describes a technique to expand and upgrade a model's tokenizer without retraining.
Liquid AI proposes Final Token Preference Optimization to reduce doom loops in language models.
Liquid AI released LFM2.5-230M, a model designed to run on diverse hardware platforms.
Liquid AI releases LFM2.5 Retrievers, bi-directional language foundation models for fast multilingual search.
Liquid AI releases LFM2.5-8B-A1B, an improved on-device mixture of experts model.
Liquid AI releases LFM2.5-VL-450M structured visual model for edge and cloud deployment.
LFM2.5-350M is a language model from Liquid AI.
LFM2-24B-A2B model runs tool-calling agents on consumer hardware without cloud infrastructure.
Liquid AI scaled up the LFM2 architecture to 24 billion parameters.
Liquid AI releases LFM2.5-1.2B-Thinking, a reasoning model under 1GB for on-device use.
LFM2.5 is next-generation on-device AI.
Liquid Labs defines frontier of intelligence through innovation.
Liquid AI advanced open science in AI with LFM2.
One pattern, one tradeoff, one production failure story. A short weekly briefing for people building agentic systems.
Weekly email, one-click unsubscribe. We only use your address to send the briefing.
