Liquid AI releases ultra-lightweight model LFM2.5-350M, specifically designed for edge device autonomous agents

MeNews · 2026-04-05T19:43:25+00:00

Liquid AI has launched the LFM2.5-350M model, an ultra-lightweight foundational model suitable for edge devices. With only 350 million parameters, its quantized size is less than 500MB. It is specifically designed for intelligent agent loops and tool invocation in low-resource environments, aiming to improve data extraction and tool utilization reliability.

MeNews

2026-04-05 19:43:25

Abstract generation in progress

ME News message, April 1 (UTC+8), Liquid AI recently released the LFM2.5-350M model, which is an ultra-lightweight foundation model specifically designed for edge devices. The model has only 350 million parameters; after quantization, its storage footprint is less than 500MB. It is intended to run autonomous agent loops and tool calls directly on edge devices with constrained compute, memory, and latency. Although its scale is relatively small, the article’s viewpoint is that the model has been specifically trained for reliable data extraction and tool use, whereas models of this typical parameter scale generally perform poorly on these tasks. (Source: InFoQ)

View Original

This page may contain third-party content, which is provided for information purposes only (not representations/warranties) and should not be considered as an endorsement of its views by Gate, nor as financial or professional advice. See Disclaimer for details.

2 Likes