Mini Compute

Mini Compute runs open-source AI language models directly on your iPhone — no account, no subscription, and no internet connection required after you download a model. Your conversations never leave your device. WHY MINI COMPUTE • Truly private — every prompt and response is processed on-device. Nothing is sent to a server, ever. • Works offline — once a model is downloaded, chat anywhere, even in airplane mode. • No account, no sign-up, no tracking — open the app and start. • Fast — models run on Apple's Metal GPU with live token streaming. CHOOSE YOUR MODEL • Start in one tap with a model recommended for your exact device. • Pick from a curated list including Llama 3.2, Qwen2.5, Gemma 2, and Phi-3.5. • Bring your own — add any compatible GGUF model from Hugging Face. • Clear "will this fit?" guidance so you only download models your iPhone can run comfortably. CHAT WITH YOUR FILES (On-Device RAG) • Attach a plain-text file and ask questions about it. • On-device retrieval finds the relevant passages — again, without anything leaving your phone. SEE HOW IT PERFORMS • A built-in Stats dashboard shows tokens per second, memory use, storage, and your conversation history at a glance. BUILT FOR IPHONE • Native SwiftUI app, designed for iOS. • Manage multiple conversations, load and unload models to free memory, and tune generation settings. Mini Compute is for anyone who wants the usefulness of an AI assistant with the privacy of keeping everything on their own device. Note: AI-generated responses can be inaccurate. Don't rely on them for medical, legal, financial, or other important decisions. Downloading models requires an internet connection; running them does not.