MiniCPM-V is OpenBMB's series of open models that understand images and video and are built to run on phones and other local devices.
In plain words
The latest version, MiniCPM-V 4.7, comes in a 35B-A3B size and a 1B size, and the larger one is the series' first mixture-of-experts model. OpenBMB says it generates more than 13 tokens a second on an iPhone 17 Pro Max. It runs on iOS, Android and HarmonyOS and works with tools such as vLLM, llama.cpp and Ollama.
A quick example
Imagine a phone app that reads a photo of a printed form and fills in the fields for you, without sending the picture to a server. A small vision model running on the phone itself makes that possible.
Where you'll meet it
- Image, document and video understanding on phones and laptops.
- Local apps built with tools such as llama.cpp and Ollama.
Related
Sources
- DocsMiniCPM-V-4.7-35B-A3B model card, OpenBMB on Hugging Face · read 12 Oct 2026
- DocsMiniCPM-V & MiniCPM-o README, OpenBMB on GitHub · read 12 Oct 2026
Checked by WhatIsAI against the sources above · 12 Oct 2026. A full five-minute explainer with a diagram may follow. Spotted an error? Report it.