Want to deploy multiple major AI big models locally on your phone? Today, we bring you a detailed introduction and demo!
Supported models include: DeepSeek R1, Llama 3.3, Qwen 2.5 and Gemma 3. These models not only support Chinese comprehension and generation, but also run smoothly on iOS and Android devices, realizing a true offline intelligence experience.
Taking the open source project PocketPal-AI as an example, we will take you step by step through the installation and configuration process, showing how to use your phone to easily invoke these powerful models for conversations and Q&A. No need to connect to the cloud, protect privacy while dramatically reducing latency, this is a practical solution that AI enthusiasts and developers can't afford to miss!
Deployment is very easy, no setup required, just click your fingers to download and install!
iOS version download: [Click to go]
Download for Android: [Click to go]
APK installation package download: [Click to go], if you are not overseas, choose 👉 [Cloud Storage Download]
PocketPal AI
Open Source ProjectOfficial Repository]
hallmark
- Offline AI Assist: Run the language model directly on your device without an Internet connection.
- Model Flexibility: Download and exchange between multiple SLMs, including DeepSeek R1, Danube 2 and 3, Phi, Gemma 2, and Qwen.
- Automatic unloading/loading: Memory is automatically managed through an offload model when the application is running in the background.
- inference setting: Customize model parameters such as system prompts, temperatures, BOS tokens, and chat templates.
- Real-time performance indicators: View the number of tokens per second and the number of milliseconds per token during AI response generation.
The latest version of PocketPal AI is now integrated with the Hugging Face Model Hub! It's very easy to browse, download and run models directly from the Hugging Face Hub within the app, and offline AI big models locally on your phone! You can choose the AI model that suits you according to the size of your phone's configuration!
