Dhruva AI: Offline AI Chat
Teen
info
0+
Downloads
About this app
Dhruva AI runs large language models directly on your phone. Every reply is generated by your own device — there is no server in the loop, no account to create, and nothing to sign in to.
The only time Dhruva touches the network is when you choose to download a model. After that you can switch on airplane mode and keep chatting.
WHAT YOU CAN DO
• Offline chat — Stream replies from any quantized GGUF model. Conversations are stored on your device.
• Reasoning you can see — Models that think before answering show their reasoning as a collapsible step instead of dumping it into the reply.
• A curated model catalog — Hand-picked models verified to run on real phones, each labelled with its download size and whether your device can handle it comfortably. Or search all of Hugging Face and pick your own.
• Characters — Build personas with their own system prompts, and keep a separate conversation with each one.
• Voice, both directions — Speech-to-text with Whisper, including Hindi and Hinglish, and text-to-speech with Piper voices. Both run on-device.
• Vision — Ask questions about a photo or a screenshot using an on-device vision model.
• Playground — Send the same prompt to two models side by side and compare the answers and the speed.
PRIVACY, CONCRETELY
• All inference happens on your device.
• No analytics, no crash reporters, no tracking SDKs, no ads.
• No account, no sign-in, no profile.
• Your conversations never leave your phone.
• Apache-2.0 licensed and fully open source, so every claim above is auditable rather than just asserted.
WHAT YOUR PHONE NEEDS
Dhruva needs a 64-bit (arm64) device running Android 8.0 or newer. Language models are memory-hungry, so as a rough guide:
• Around 1B parameters — 4 GB RAM or more
• 3B to 4B parameters — 6 GB RAM or more
• 7B parameters and up — 8 GB RAM or more
Smaller models run comfortably on most modern phones; larger ones will be slower. Models are downloaded separately and use storage, so check the size shown next to each one before downloading.
A NOTE ON THE NAME
ध्रुव (Dhruva) is the pole star, the one fixed point in the night sky that never moves. This app is independent and is not affiliated with AI4Bharat's Dhruva speech platform.
Source code: https://github.com/AnshRajput/dhruva-app
The only time Dhruva touches the network is when you choose to download a model. After that you can switch on airplane mode and keep chatting.
WHAT YOU CAN DO
• Offline chat — Stream replies from any quantized GGUF model. Conversations are stored on your device.
• Reasoning you can see — Models that think before answering show their reasoning as a collapsible step instead of dumping it into the reply.
• A curated model catalog — Hand-picked models verified to run on real phones, each labelled with its download size and whether your device can handle it comfortably. Or search all of Hugging Face and pick your own.
• Characters — Build personas with their own system prompts, and keep a separate conversation with each one.
• Voice, both directions — Speech-to-text with Whisper, including Hindi and Hinglish, and text-to-speech with Piper voices. Both run on-device.
• Vision — Ask questions about a photo or a screenshot using an on-device vision model.
• Playground — Send the same prompt to two models side by side and compare the answers and the speed.
PRIVACY, CONCRETELY
• All inference happens on your device.
• No analytics, no crash reporters, no tracking SDKs, no ads.
• No account, no sign-in, no profile.
• Your conversations never leave your phone.
• Apache-2.0 licensed and fully open source, so every claim above is auditable rather than just asserted.
WHAT YOUR PHONE NEEDS
Dhruva needs a 64-bit (arm64) device running Android 8.0 or newer. Language models are memory-hungry, so as a rough guide:
• Around 1B parameters — 4 GB RAM or more
• 3B to 4B parameters — 6 GB RAM or more
• 7B parameters and up — 8 GB RAM or more
Smaller models run comfortably on most modern phones; larger ones will be slower. Models are downloaded separately and use storage, so check the size shown next to each one before downloading.
A NOTE ON THE NAME
ध्रुव (Dhruva) is the pole star, the one fixed point in the night sky that never moves. This app is independent and is not affiliated with AI4Bharat's Dhruva speech platform.
Source code: https://github.com/AnshRajput/dhruva-app
Chat with AI models that run entirely on your phone. No cloud, no account.
Updated on
Featured stories
Safety starts with understanding how developers collect and share your data. Data privacy and security practices may vary based on your use, region, and age. The developer provided this information and may update it over time.
- No data shared with third partiesLearn more about how developers declare sharing
- No data collectedLearn more about how developers declare collection
What’s new
First public release of Dhruva AI.
- Offline chat with language models running entirely on your device
- Curated catalog of models verified to run on real phones
- Characters with their own system prompts
- Voice in both directions, plus vision for photos and screenshots
- Playground to compare two models side by side
No account, no telemetry, no cloud inference.
- Offline chat with language models running entirely on your device
- Curated catalog of models verified to run on real phones
- Characters with their own system prompts
- Voice in both directions, plus vision for photos and screenshots
- Playground to compare two models side by side
No account, no telemetry, no cloud inference.
App support
About the developer
MANTRA INFOTECH
connect@mantrainfotech.in
9\836, Extension \1, Shalimar Garden
Ghaziabad, Uttar Pradesh 201005
India
+91 70113 40142
