OpenBMB/MiniCPM-V-Apps

MiniCPM-V apps — fully offline multimodal chat on iOS / Android / HarmonyOS

What it solves

This project provides a set of mobile application demos that allow the MiniCPM-V family of multimodal models to run fully on-device, eliminating the need for cloud-based inference. It supports multiple mobile operating systems, including iOS, Android, and HarmonyOS NEXT.

How it works

The applications leverage llama.cpp-omni (integrated as a git submodule) to execute GGUF-formatted model weights locally on mobile hardware. The project includes separate build pipelines for each platform: an Xcode project for iOS, a Gradle/Kotlin project for Android, and a DevEco Studio project for HarmonyOS. Users can either download pre-built packages or build from source and load specific model versions (such as MiniCPM-V 2.6, 4.0, 4.6, or the text-only MiniCPM5-1B).

Who it’s for

  • Mobile Developers: Those looking to integrate local LLMs and multimodal models into mobile apps.
  • AI Researchers: Users wanting to test the performance of multimodal models on edge devices.
  • End Users: People who want to run private, offline multimodal AI on their smartphones.

Highlights

  • Cross-Platform Support: Native demos for iOS, Android, and HarmonyOS.
  • Multimodal Capabilities: Supports vision-language models (MiniCPM-V) and text-to-speech (VoxCPM2).
  • On-Device Execution: Runs fully locally via llama.cpp, ensuring privacy and offline availability.
  • Hardware-Specific Guidance: Provides detailed RAM requirements for different model quantizations to ensure smooth performance on various devices.

Related

  • Project
    OpenBMB/MiniCPM-V

    A series of pocket-sized multimodal LLMs designed for efficient on-device deployment, enabling high-performance image, video, and real-time omnimodal interaction on mobile phones.

  • Project
    OpenSQZ/MiniCPM-V-CookBook

    A comprehensive implementation guide and deployment cookbook for the MiniCPM series, enabling the integration of vision, speech, and text AI capabilities across edge devices and enterprise servers.

  • Project
    qualcomm/GenieX

    An on-device Gen AI inference runtime for Qualcomm Snapdragon devices that allows users to run LLMs and VLMs locally using GGUF models or optimized Qualcomm AI Hub bundles.

  • Project
    nobodywho-ooo/nobodywho

    An on-device inference engine that enables developers to run LLMs, speech-to-text, and text-to-speech locally and offline across multiple platforms.

  • Project
    ggml-org/llama.cpp

    A high-performance C/C++ implementation for local LLM and VLM inference, supporting a wide range of hardware through quantization and hardware acceleration.