Private, offline, and powerful. Chat with Qwen 3.5, Gemma 4, and vision models that understand your photos — no internet connection needed. Your data stays on your device.
Everything runs on your device. No data sent to servers. No tracking. Your conversations stay yours.
No internet required. Chat with AI models anywhere, anytime. Perfect for travel or areas with poor connectivity.
Run Qwen 3.5, Google Gemma 4, GLM 4.6V, and Ministral with GPU acceleration — or add any GGUF model by URL. Small and fast to flagship-class.
Every catalog model supports vision. Snap a photo or attach one and ask about it — read documents, identify objects, describe scenes.
Adjust temperature, top-k, top-p, and more. Enable thinking mode to watch step-by-step reasoning before the answer.
Download models once and use them forever. No subscriptions. No API costs. Truly yours.
Model sizes range from about 0.8GB to 6GB. Qwen 3.5 0.8B is the smallest and fastest, while larger models like Qwen 3.5 9B and GLM 4.6V give better answers but need more storage and RAM.
No. All processing happens entirely on your device. Your conversations never leave your phone. There are no servers, no tracking, and no data collection.
The app works on iPhones and Android devices, with GPU acceleration on Metal-capable iPhones and Snapdragon Adreno phones. We recommend 4GB+ RAM; the largest models want 8GB. The app shows a "fits your device" badge on every model.
Yes! Once you download a model, it works completely offline. Perfect for travel, remote areas, or when you want complete privacy.
Qwen 3.5 is the best all-rounder, Gemma 4 is efficient on everyday phones, GLM 4.6V Flash is flagship-class vision, and Ministral is a compact alternative. Every catalog model can understand images, and larger models give better answers but run slower.
Yes, the app is completely free. Download models once and use them forever. No subscriptions, no API costs, no hidden fees.
Have questions, feedback, or need support? We'd love to hear from you!
[email protected]We typically respond within 24-48 hours.