LiquidAI Unveils Lightweight AI Models for Edge Devices
LiquidAI Introduces Liquid Nanos Series for Edge Computing
September 29, 2025 — LiquidAI has officially launched its Liquid Nanos series of lightweight AI models, specifically designed for edge computing devices. These models can run efficiently on small hardware such as Raspberry Pi, making advanced AI capabilities accessible without requiring powerful computing resources.

Model Specifications and Features
The Liquid Nanos series offers two parameter versions:
- 350M parameters: Ideal for low-power devices
- 1.2B parameters: For slightly more demanding applications
All models support the GGUF quantization format, optimizing performance while minimizing resource utilization. This makes them particularly suitable for deployment scenarios where computational power and energy efficiency are critical considerations.
Application Scenarios
The series covers five primary use cases:
- Translation: Professional-grade language conversion
- Extraction: Data parsing and information retrieval
- RAG (Retrieval-Augmented Generation): Enhanced question answering systems
- Tool Calling: API integration and function execution
- Mathematical Reasoning: Complex problem-solving capabilities
Available Models on Hugging Face
The first twelve task-specific models are now available on the Hugging Face platform:
- LFM2-350M-ENJP-MT: Japanese-English translation model
- LFM2-350M/1.2B-Extract: Data extraction models
- LFM2-1.2B-RAG: Retrieval-augmented generation model
- LFM2-1.2B-Tool: Tool calling model
- LFM2-350M-Math: Mathematical reasoning model
The complete collection can be accessed at: https://huggingface.co/collections/LiquidAI/liquid-nanos-68b98d898414dd94d4d5f99a
Industry Impact
The release represents a significant advancement in bringing sophisticated AI capabilities to edge devices. Developers can now implement complex AI functions directly on local hardware without relying on cloud-based solutions.
The lightweight nature of these models opens new possibilities for:
- IoT device intelligence
- Real-time processing applications
- Privacy-sensitive computations where data cannot leave the device As edge computing continues to evolve, the Liquid Nanos series positions itself as a versatile toolkit enabling innovation across multiple industries. --- ### Key Points: ✅ Designed specifically for edge computing devices like Raspberry Pi ✅ Available in two sizes: 350M and 1.2B parameter versions ✅ Supports GGUF quantization format for optimized performance ✅ Twelve specialized models now available on Hugging Face ✅ Covers five core application scenarios including translation and RAG