10月24日 的 Offline AI Chat: Private 應用程式分析

Offline AI chat on your phone — no internet needed once a model is downloaded. Offline AI Chat is an on-device AI assistant and private AI chatbot for Android. Choose and download a local AI model, then use it for everyday questions, text generation, rewriting, summaries, translation, studying, brainstorming, math, coding, and focused work without relying on cloud AI inference. Your prompts and generated answers are processed by the selected model on your phone. WHAT TO EXPECT The AI runs entirely on your phone, so its models are smaller than online chatbots. On entry-level phones answers can be slower and less accurate. The first model download needs internet and free storage (from about 470 MB to several GB). Start with the small Tiny model, then try a larger one if your phone can handle it. WRITE, LEARN, AND ORGANIZE Turn a rough idea into an outline, polish an email, simplify a paragraph, compare options, create a study guide, explain a concept, or prepare notes for a project. Ask follow-up questions in the same chat and keep useful conversations in a searchable local history. Use it as a writing assistant, then attach supported documents, photos, or audio when you need to give a compatible model more context. Local models differ in what they can understand, so availability and answer quality depend on the model you install and the capabilities of your phone. CHOOSE A MODEL FOR YOUR DEVICE Browse compact models for a quicker start or larger models for more demanding tasks. The optional device check can recommend a suitable choice based on your phone; you can skip the recommendation and select another compatible model yourself. Model downloads are substantial and additional free storage is required. After installation, model preparation and AI inference run locally. Available local AI models: • Tiny v2 (Granite 4.0 350M) — ~468 MB • Quick (Gemma 4 E2B) — ~2.6 GB • Smart (Gemma 4 E4B) — ~3.7 GB • Compact (Phi-4 Mini) — ~3.8 GB Premium models: Gemma 4 12B, Qwen2-VL-2B, DeepSeek R1 1.5B, Qwen2.5 1.5B, Qwen2.5 Coder 3B, Qwen3 0.6B, FLUX.2. FREE AND PREMIUM FEATURES Saved chat history remains available without a subscription. Premium unlocks additional supported models and features, including questions about text-based PDF documents, custom system prompts, chat export, compatible local-model import, and private on-device image generation and editing where supported. Scanned or password-protected PDFs are not supported. Premium availability and feature support can vary with the model and device. WORK WITH CONTEXT Use document chat for text-based PDF files, ask about an image with a vision-capable model, or discuss an audio attachment when the selected model and device support it. Features depend on your hardware and the selected local model. After a model is downloaded, AI chat and inference work offline. Internet is still needed for app updates, model downloads, Google Play purchases or restores, and optional diagnostics when connected. PRIVATE BY DESIGN AI conversation content is handled locally for inference rather than sent to a cloud AI service to generate an answer. This app does not promise current web information: local models can make mistakes, so check important facts and professional advice. Local AI technology, CPU, and GPU: Backend: Google LiteRT / LiteRT-LM · Acceleration: CPU / GPU Local LLM inference uses the Google LiteRT and LiteRT-LM backend. Auto uses CPU by default for stability. Compatible phones can use GPU acceleration when GPU is selected or when experimental GPU probing is enabled. Image analysis requires a compatible vision model and a working GPU runtime. Speed, memory use, battery impact, and feature support depend on the model and device. Download sizes are approximate, and additional free storage is required.