GitLab Transcend - Intelligent orchestration
Mini toolsLFM2.5 350M
By Liquid AI
LFM2.5-350M is an upgraded 350M-parameter model built on Liquid AI’s LFM2 architecture, with additional pretraining scaled from 10T to 28T tokens plus large-scale reinforcement learning. It is aimed at efficient edge and production deployment, with support across engines such as llama.cpp, MLX, vLLM, SGLang, ONNX, and OpenVINO. Liquid AI presents it as especially strong for instruction following, tool use, and data extraction, including large gains over LFM2-350M on IFBench, CaseReportBench, and BFCLv3, while explicitly noting it is not recommended for math, code, or creative writing.
New
Text
Gen 7
Overview
LFM2.5-350M is Liquid AI’s compact 350M-parameter language model optimized for fast inference, tool use, data extraction, and structured outputs. It is designed to run across cloud GPUs, CPUs, and edge devices, and is positioned for lightweight agentic workflows and large-scale data processing rather than math, coding, or creative writing.
About Liquid AI
Liquid AI is an MIT spin-off building efficient general-purpose AI models (Liquid Foundation Models, or LFMs) that run on edge devices with less memory and power.
They recently raised $250M in Series A funding to scale model development and deployment.
Tools using LFM2.5 350M
No tools found for this model yet.
Last updated: April 1, 2026
