LiquidAI/LFM2-700M

LFM2-700M: A Hybrid Model for Efficient Edge AI

LFM2-700M is part of Liquid AI's new generation of hybrid models, specifically engineered for high performance on edge devices. This 0.7 billion parameter model introduces a novel architecture combining multiplicative gates and short convolutions, enabling significant improvements in speed and memory efficiency.

Key Capabilities & Features

• Optimized Performance: Achieves 3x faster training and 2x faster decode/prefill speeds on CPU compared to its previous generation and Qwen3, respectively.
• Superior Benchmarking: Outperforms similarly-sized models across various categories, including knowledge, mathematics, instruction following, and multilingual tasks.
• Flexible Deployment: Designed for efficient operation on CPU, GPU, and NPU hardware, supporting deployment on smartphones, laptops, and vehicles.
• Multilingual Support: Supports English, Arabic, Chinese, French, German, Japanese, Korean, and Spanish.
• Tool Use: Features a structured tool-use mechanism with JSON function definitions and Pythonic function calls.
• Training: Utilizes knowledge distillation from LFM1-7B, large-scale SFT, custom DPO, and iterative model merging.

Recommended Use Cases

LFM2-700M is particularly well-suited for fine-tuning on narrow use cases to maximize performance. It is recommended for:

• Agentic tasks
• Data extraction
• Retrieval-Augmented Generation (RAG)
• Creative writing
• Multi-turn conversations

However, it is not recommended for knowledge-intensive tasks or those requiring advanced programming skills.