Local LLM
by Md Arafath Rahman
Run open-weight Large Language Models directly on your Android phone with zero internet connection, zero tracking, and complete privacy. Integrated with llama.cpp, custom TTS voices (Kokoro, Kitten, Supertonic), offline hardware benchmarks, and dynamic Material 3 Expressive theming.
App Screens
Explore all 17 core interfaces of Local LLM. Seamlessly switch between light and dark themes to examine chat views, model catalogs, and benchmark diagnostics.
Pocket Intelligence, Zero Cloud
Uncompromising on-device AI execution with complete offline autonomy and whisper-fast neural inference.
llama.cpp Engine
Direct C++ native neural inference optimized for ARM NEON and Vulkan GPU acceleration on mobile chipsets.
Offline TTS Voices
Natural, high-fidelity neural speech synthesis using Kokoro, Kitten, and Supertonic models without internet.
Integrated Model Hub
Browse, download, and initialize quantized GGUF weights including Gemma 4, LFM, and Ministral directly in-app.
Hardware Diagnostics
Track tokens per second, memory bandwidth, temperature, and thermal throttling in real time.
Material 3 Expressive
Dynamic palette tokenization, multi-accent picking, fluid animations, and high-readability typography.
100% Offline Privacy
No telemetry, zero cloud calls, no accounts, and no data leaving your phone. Everything stays in your pocket.
Color Palette
High-contrast modern violet and deep dark palette crafted for comfortable night coding and bright daylight readability.
Dark Theme
Light Theme
Built With
Android (Kotlin)
Material 3 Expressive
Jetpack Compose
Room Database
StateFlow & Coroutines
llama.cpp Native
GNU GPLv3 Open Source
Interested in this project?
Local LLM is open source. Check out the source code on GitHub, fork it, or build it locally with Android Studio.