LlamaPal - Private Offline AI

AI you own, not AI you rent.

LlamaPal runs powerful open-source AI models directly on your Android phone: completely offline, fully private, no signup. Once a model is downloaded, your conversations never leave your device. No internet required. No cloud servers. No accounts. Ever.

⚡ WHY LLAMAPAL?

• 100% Offline AI: chat without internet after the model downloads. Works in airplane mode, on flights, in dead zones.
• Private by design: no accounts, no telemetry of conversations, no cloud sync. Inference runs on your device using llama.cpp.
• Free, with optional Pro: most features unlocked forever. Lightweight ads keep the lights on. Pro removes ads if you prefer.
• Choose your model: browse a curated catalog of open-source LLMs (Gemma 4, Qwen 3.5, Ministral 3, Granite 4.2, and more). Each model is labeled by speed and quality so you can match it to your device.
• Voice conversations: tap the voice button and just talk. LlamaPal listens, thinks and answers out loud, hands-free, in a choice of voices. Speech recognition and speech synthesis both run on your phone.
• Built-in characters: chat with pre-made AI personas, or build your own with Pro.
• Voice input: dictate your prompts with on-device speech recognition.
• Modern dark UI: warm, hand-crafted Material 3 design built for long sessions.

🛡 PRIVACY YOU CAN VERIFY

LlamaPal does not collect chat data. The on-device LLM runtime is open-source so you can inspect what's running on your phone. The only network traffic is: catalog browsing, model downloads from Hugging Face, anonymous crash reports (Firebase Crashlytics), and ads (AdMob). Everything else (chat history, characters, settings, personas) stays local.

📚 ANY OPEN-SOURCE MODEL, ON YOUR HARDWARE

Browse community-tested GGUF models including Gemma 4, Qwen 3.5, Ministral 3, Granite 4.2, LFM2.5 and MiniCPM 5. Pick the size that fits your device: small models for instant replies, larger models for deeper reasoning. New models are added regularly via a remote catalog.

💎 LLAMAPAL PRO (OPTIONAL)

Pro is for users who want to support development and skip the ads. Includes:
• No ads anywhere in the app
• Unlimited model downloads
• Unlimited voice talk time
• Custom themes and app icons

Available as a monthly or annual subscription. Pricing is localized for your region. See the in-app store for details. Cancel anytime in Google Play.

Core chat is always free, and voice conversations come with free talk time every day. We will never paywall the AI itself.

📱 REQUIREMENTS

• Android 9 (API 28) or higher
• 64-bit ARM device (arm64-v8a)
• 6 GB+ RAM recommended for most models
• 4–8 GB free storage per model

🦙 OPEN SOURCE INFERENCE

The on-device LLM runtime in LlamaPal will be released as a standalone open-source library on Maven Central. You will be able to inspect, audit, and reuse the exact code that runs your conversations.

LlamaPal is a private offline AI chatbot, a ChatGPT offline alternative for anyone who wants AI without internet, AI without accounts, and full ownership of their data.

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

Recent changes:
New in 1.1.2
• Models too large for your phone now show Won't Fit instead of freezing the app, and an oversized download is no longer loaded at launch.
• Fixed a freeze in the middle of a reply on some phones.
• Fixed a crash from the ad consent screen and one from the Play purchase screen.
• Stopping a download is now immediate, and downloads paused by Android resume where they left off.