LlamaPal - Private Offline AI

Offline AI chat that runs 100% on your phone. No internet, no account, no cloud.

LlamaPal is a private AI assistant powered by open-source models (local LLMs) that you download once and run on your own device. After the download, nothing you type or say ever leaves your phone. Chat on a flight, in the metro, in a dead zone, or with Wi-Fi switched off: it just works.

⚡ WHY LLAMAPAL

• Works offline: AI chat without internet, anywhere. Airplane mode is fine.
• Private by design: no account, no sign-up, no chat data sent to any server. Inference runs on-device with llama.cpp.
• Talk to it: tap the voice button and have a hands-free voice conversation. Speech recognition and speech synthesis run on your phone too.
• Show it things: add photos and files to a chat. Every model reads the words in a photo, and vision models can see the picture itself.
• Pick your model: a curated catalog of open-source LLMs including Gemma 4, Qwen 3.5, Ministral 3, Granite 4.2, LFM2.5 and MiniCPM 5, each labeled by speed and quality so you can match it to your device.
• Built-in characters: chat with ready-made AI personas for coding, writing, tutoring and casual conversation.
• Free to use: core chat is free forever. Light ads keep the lights on, and optional Pro removes them.

🛡 PRIVACY YOU CAN CHECK

LlamaPal does not collect your conversations. Chats, characters and settings are stored only on your phone. The only network traffic is the model catalog, model downloads from Hugging Face, anonymous crash reports and diagnostics, and ads. The on-device inference engine is open-source, so you can see exactly what runs on your device.

📚 ANY OPEN-SOURCE MODEL, YOUR HARDWARE

Browse community-tested GGUF models and pick the size that fits your phone: small models for instant replies, larger models for deeper reasoning and coding. New models are added regularly without an app update.

🎙 VOICE CONVERSATIONS

Just talk. LlamaPal listens, thinks and answers out loud in a choice of voices, fully offline. Dictate prompts with on-device speech recognition, or have a full back-and-forth conversation hands-free.

📎 PHOTOS AND FILES

Snap a receipt, pick a screenshot or attach a PDF, Word document, CSV or code file, then ask about it. Every model reads the words in a photo with on-device text recognition. Gemma 4, Qwen 3.5 and Ministral 3 models can see the picture itself after a one-time vision add-on download. Photos and files are read on your phone and never uploaded.

💎 LLAMAPAL PRO (OPTIONAL)

Pro supports development and removes ads. It includes:
• No ads anywhere in the app
• Unlimited model downloads
• Unlimited voice talk time
• Unlimited photos and files
• Custom themes and app icons

Monthly or annual, priced for your region, cancel anytime in Google Play. Core chat is always free and voice comes with free talk time every day. The AI itself is never paywalled.

📱 REQUIREMENTS

• Android 9 or higher, 64-bit (arm64) device
• 6 GB+ RAM recommended for most models
• 2 to 8 GB free storage per model

🦙 OPEN SOURCE

The LlamaPal inference engine (the llama.cpp runtime that powers every chat) will be published as a standalone open-source library on Maven Central.

Looking for a ChatGPT offline alternative, a local AI chat for Android, or a private LLM that works without internet? That is LlamaPal: offline AI chat, local models, zero cloud.

Questions or bug reports: dev.llamapal@gmail.com

Recent changes:
New in 1.2.0
• Add photos and files to a chat: PDF, Word, text, CSV and code. Everything is read on your phone and nothing is uploaded.
• Every model reads the words in a photo. Gemma 4, Qwen 3.5 and Ministral 3 can see the picture after a one-time add-on download.
• New Photos filter in the Model Hub.
• Fixed Download stuck on "No internet" and chats out of order after a clock change.