OpenBMB, in collaboration with Tsinghua University, has open-sourced MiniCPM-V 4.6, a multimodal large language model with 1.3 billion parameters. The model is designed to run efficiently on a single NVIDIA RTX 4090 GPU, making advanced AI accessible to individual developers and small teams.
According to the project's GitHub repository and official announcements, MiniCPM-V 4.6 achieves performance comparable to larger models on benchmarks such as MMMU and MathVista. It supports image and text inputs, enabling tasks like visual question answering and document analysis.
The release includes pre-trained weights and inference code under an open-source license. The model's small size allows for local deployment without cloud dependencies, addressing privacy and latency concerns. As of May 2026, the project has gained attention in the AI community for its efficiency and accessibility.
