Liquid AI Releases LFM2.5-VL-3B: A 3B Vision-Language Model for On-Device Screen and Document Understanding

Liquid AI has released LFM2.5-VL-3B, a 3.1-billion-parameter vision-language model optimized for on-device deployment across mobile, web, and desktop platforms. The model reads digital screens, anchors objects to screen coordinates, analyzes documents and charts, and executes tool calls from visual input while maintaining low latency. The model scores an average of 69.4 across 28 vision benchmarks, matching the performance of larger 4.7-billion-parameter competitors. It operates within approximately 3 GB of memory and decodes at 228 tokens per second on an Apple M5 Max processor. The 32,768-token context window supports 16 languages, and weights are available in native, GGUF, ONNX, and MLX formats compatible with leading inference engines including vLLM, SGLang, and llama.cpp.