OpenPocket

Run large language models on your iPhone, fully offline and private. Supports Gemma, Qwen, DeepSeek and more — all inference happens on-device, no internet required.

Open Pocket is an intelligent assistant that runs large language models entirely on your device.

Privacy First: No servers, no data uploads — your conversations belong to you and only you. All inference runs locally, no internet connection needed,and your data never leaves your device.

Key Features:

• Supports Llama, Qwen, DeepSeek, Gemma, Phi, Mistral and other popular open-source models
• Powered by Apple MLX framework, fully utilizing Apple Silicon performance
• One-tap model download from HuggingFace and ModelScope
• Streaming conversations with real-time inference output
• Chain-of-thought models (DeepSeek R1, Qwen 3) with visible reasoning process
• Multi-session management with full conversation history
• Built-in performance benchmarks to see real model speed on your device
• Completely free, no subscriptions, no ads