Liquid AI Releases LFM2.5-2.6B: An On-Device Agentic Model With 128K Context, Tool Calling, And Open Weights

Liquid AI has introduced LFM2.5-2.6B, an open-weight language model optimized for local execution. The architecture combines convolution blocks with grouped query attention across 30 layers to enable agentic capabilities without cloud dependency.

Key specifications include a 2.69B parameter count and a context window of 131,072 tokens. The model is built to handle tool calling and multi-step planning directly on consumer hardware, addressing latency and privacy concerns associated with cloud-based inference.

This release targets the growing demand for efficient on-device AI solutions. By providing open weights, Liquid AI aims to allow developers to integrate agentic workflows into local applications while maintaining performance comparable to larger cloud-based counterparts.

This page provides an editorial summary based on publicly available information. It is not a republished article. Use the source link below for the original report.