What is LFM2-VL?
LFM2-VL is Liquid AI’s new series of vision-language foundation models designed for efficient, on-device deployment. It supports both text and image inputs with native resolution up to 512×512, enabling high-performance multimodal understanding. With up to 2× faster inference speed on GPUs, LFM2-VL delivers lightweight yet powerful solutions for devices ranging from phones to single-GPU systems, balancing accuracy and efficiency for real-world applications.
Features
- 2× faster inference speed on GPUs
- Supports text and image inputs
- Native resolution processing up to 512×512
- Tunable speed-quality tradeoffs at inference
- Lightweight variants for resource-limited devices
Use Cases
- Image captioning on mobile devices
- Multimodal search across text and visuals
- Edge deployment for wearables
- Fast OCR and document analysis
