88Z AI, 12 May — OpenBMB releases MiniCPM-V 4.6, an edge multimodal large model with 1.3B parameters and ~1.6GB size, supporting pure local offline deployment on iOS, Android, and HarmonyOS NEXT.
At a glance
- Model size is ~1.6 GB, running smoothly with a minimum of 6 GB RAM
- Supports pure local offline deployment on iOS, Android, and HarmonyOS NEXT via llama.cpp
The update
OpenBMB has officially open-sourced MiniCPM-V 4.6, an edge-side multimodal large model. The language component features only 1.3 billion parameters, with a total model size of approximately 1.6 GB. It has low hardware requirements, running smoothly with a minimum of 6 GB of RAM. Built on the llama.cpp framework, the model fully supports pure local offline deployment across three major mobile platforms: iOS, Android, and HarmonyOS NEXT. This allows users to operate the model without relying on network connectivity. In terms of capabilities, MiniCPM-V 4.6 offers robust multimodal processing, including image understanding, optical character recognition (OCR), and video understanding, aiming to provide efficient and privacy-safe intelligent interaction for mobile devices.
Sources & editorial note
This article is an AI-assisted synthesis by 88Z AI based on the public sources below. Product claims and availability are attributed to their publishers. It does not represent independent testing or first-hand reporting.
Source dates above refer to the original announcements; the date under the headline is this article’s publication on 88Z AI.