Solo is a private AI assistant that runs entirely on your phone. There is no cloud, no account, and no chat server. Your messages stay on your device.
Why Solo is different
• Truly on-device: the language model runs locally using llama.cpp. Once the model file is downloaded on first launch, the app works fully offline — even in airplane mode.
• No backend for chat: your prompts and responses never travel to a server. Nothing is uploaded, nothing is logged remotely.
• No account required: open the app and start chatting. No sign-up, no email, no phone number.
• Your data, your device: chat history is stored only in this app's private storage. You can clear it any time from Settings.
What you can do with Solo
• Ask questions, brainstorm, draft emails, summarise text, get explanations.
• Use it on flights, on the subway, in low-signal areas — it doesn't need a connection.
• Stop a response mid-stream when you've seen enough.
• Long-press any message to copy it.
• Manage installed models and free up space from Settings.
How it works
On first launch, Solo downloads a compact open-source language model (~400 MB, Qwen2.5-0.5B-Instruct) from Hugging Face. After that, the model lives on your device and runs locally on every chat. No further network
calls are made for inference.
Performance notes
• Optimised for modern phones. Older or low-RAM devices may respond more slowly.
• Running a language model is computationally intensive — your phone may warm up during long sessions.
Privacy at a glance
• Chat content: stays on device. Never sent anywhere.
• Anonymous usage analytics: limited app-event analytics (screen views, model load events) are collected to improve reliability. No message content is ever included.
• Push notifications: optional, used only for occasional product updates if you grant permission.
Permissions
• Internet: needed once to download the model and to deliver optional push notifications.
• Storage: needed to save the downloaded model on your device.
Open source models, open standards
Solo uses widely-reviewed, open-weights models distributed under permissive licenses. You can manage and remove models at any time from the Settings screen.
Made with care by Aftab Ali — an independent developer who believes your conversations belong to you, not to a server farm.
Private on-device AI chat. The model runs on your phone. Works fully offline.