MiniCPM-V

MiniCPM-V is an open-source family of multimodal AI models developed by OpenBMB for visual understanding and efficient on-device inference. It processes images and text to answer questions, interpret documents, and analyze visual content. Its latest MiniCPM-V 4.7 release includes a compact 1B model and a 35B-A3B mixture-of-experts model. The larger model uses Lookahead Expert Scheduling to accelerate inference on memory-constrained devices. Quantized versions reduce hardware requirements, while speculative decoding further improves generation speed. The broader MiniCPM family also includes MiniCPM-o models for real-time audio, video, and speech interactions. Developers can use pretrained checkpoints, inference frameworks, and fine-tuning resources to build multimodal applications.