Liquid AI

LFM2.5-2.6B v2.5

LFM2.5-2.6B is a compact, agentic foundation model designed for efficient on-device deployment. It features a 128K context window and is specifically optimized for agentic workflows, including planning, tool use, and multi-step task execution.

LFM2.5-1.2B-Thinking v1.0

LFM2.5-1.2B-Thinking is a general-purpose text-only model designed for on-device deployment, building upon the LFM2 architecture with extended pre-training and reinforcement learning. It offers high-quality AI performance comparable to much larger models, optimized for efficient edge inference on various hardware platforms.

LFM2.5-VL-3B v2.5

LFM2.5-VL-3B is a 3.1-billion-parameter vision-language model designed for high-performance, on-device inference. It features significant improvements in screen understanding, grounding, and function calling, allowing it to outperform larger models while maintaining low latency on edge hardware.

LFM2.5 v2.5

LFM2.5 is a new generation of small foundation models built on the LFM2 architecture, optimized for on-device and edge deployments. The model family includes variants such as LFM2.5-1.2B-Base, LFM2.5-1.2B-Instruct, and extends to Japanese, vision language, and audio language models. It is released as open weights on Hugging Face and exposed through the LEAP platform.

LFM2.5-Encoder-230M v2.5

LFM2.5-Encoder-230M is a lightweight, multilingual bidirectional encoder built on the LFM2 hybrid architecture. It is designed for high-performance, low-latency tasks such as text classification, retrieval, and PII detection, optimized for on-device and browser-based deployment.

LFM2.5-Encoder-350M v2.5

LFM2.5-Encoder-350M is a multilingual bidirectional encoder model built on the LFM2 hybrid architecture. It is designed for high-performance, on-device tasks such as text classification, retrieval, and semantic similarity across 15 languages.