ErtasAI/LFM2.5-1.2B-Instruct
ErtasAI/LFM2.5-1.2B-Instruct is a 1.2 billion parameter instruction-tuned hybrid model developed by Liquid AI, designed for efficient on-device deployment with a 32,768 token context length. It features extended pre-training on 28T tokens and multi-stage reinforcement learning, enabling best-in-class performance that rivals larger models. This model excels at agentic tasks, data extraction, and RAG, offering fast edge inference on CPUs and mobile NPUs.
Loading preview...
LFM2.5-1.2B-Instruct Overview
LFM2.5-1.2B-Instruct is a 1.2 billion parameter instruction-tuned model from Liquid AI, specifically engineered for on-device deployment and efficient edge inference. It builds upon the LFM2 architecture, benefiting from significantly extended pre-training (28 trillion tokens) and large-scale multi-stage reinforcement learning. This model delivers high-quality AI performance comparable to much larger models, while maintaining a low memory footprint (under 1GB).
Key Capabilities & Features
- Optimized for On-Device: Designed for fast inference on CPUs (239 tok/s) and mobile NPUs (82 tok/s), with day-one support for llama.cpp, MLX, and vLLM.
- Compact yet Powerful: A 1.2B parameter model that rivals the performance of larger models, making high-quality AI accessible for edge devices.
- Extensive Training: Pre-trained on 28 trillion tokens with a 32,768 token context length, supporting English, Arabic, Chinese, French, German, Japanese, Korean, and Spanish.
- Tool Use Support: Features robust function calling capabilities, allowing for integration with external tools and APIs.
- Benchmark Performance: Demonstrates strong performance across various benchmarks, often outperforming other sub-2B models in categories like GPQA, MMLU-Pro, and IFEval.
Recommended Use Cases
- Agentic Tasks: Ideal for applications requiring intelligent agents.
- Data Extraction: Efficiently extracts information from text.
- Retrieval Augmented Generation (RAG): Suitable for enhancing generation with retrieved information.
- Edge Deployment: Excellent for applications on vehicles, mobile devices, laptops, IoT, and embedded systems due to its low memory usage and fast inference.