Pocket LLM for Android
Pocket LLM for Android turns an Android phone into a local LLM server for a home network, or a private on-device chat.
Verified
Core facts
Read by a script from the official API, not rewritten by a model.
- License
- MIT
- Languages
- Kotlin
- Maintainer
- dineshsoudagar
- Verified
What it is
Pocket LLM for Android turns an Android phone into a local LLM server for a home network, or chats privately on-device with text, images, PDFs, audio, and offline speech-to-text. Users can chat in the app, open its Web UI on a computer, or connect a compatible OpenAI client while the model runs on the phone. The small base APK is installed first, then a built-in model is downloaded or a LiteRT model (beta) is imported. Models are not bundled inside the APK.
Main features
The app runs models on the phone with ONNX or LiteRT, offline after installation.
Inference and chat data stay on the phone, with no telemetry.
Built-in models include Qwen2.5, Qwen3, DeepSeek R1 Distill Qwen, and Gemma 4.
Models can be downloaded, switched, and deleted inside the app.
Reasoning can be toggled on supported Qwen3 and Gemma models.
Recordings can be turned into editable text with offline multilingual Whisper, including German.
Getting started
Clone this repository.
Install the latest Android Studio.
Build and install the app on your Android device.
On first launch, choose a model from the built-in model picker.
Download the selected model directly inside the app.
For the LAN server, load a model, open LAN Server in the Android app, enter a password of at least 8 characters, and tap Save password.
pocket_llm_src/Who it is for
The app attaches text files and PDFs, including scanned PDFs read with OCR. UTF-8 or UTF-16 text files can be attached up to 16 MiB, and PDFs up to 64 MiB and 500 pages in the Android app. Android-decodable audio can be attached up to 512 MiB and 2 hours, or audio can be recorded in-app.
When native audio is unavailable, transcription uses multilingual Whisper tiny int8 after a separate download of about 100 MB. Each Android chat keeps one active document or audio source; do not mix it with images in one send. Each LiteRT model's context size can be set to balance history and memory use.
The LAN server serves the phone's selected model to devices on the same private network, and the Web UI is password-protected while inference stays on the phone. The Web UI lets users chat in a PC browser, upload PDFs or images, and save conversations on the phone. In the browser, direct image uploads require a compatible model.
The OpenAI-compatible API connects compatible clients through model and chat-completion endpoints. Custom models do not expose native tool calling through the LAN API. Custom-model compatibility is experimental, and larger models or contexts may fail to load or crash.
Stop the LAN server before changing the model or password. The requirements list a physical Android device for deployment and testing, and at least 4 GB RAM for smaller models. Before a model is loaded, Pocket LLM checks available storage and validates the downloaded files, and the app automatically attempts initialization with the available backend fallbacks.
Saved chat history is kept independently and is not deleted by model-load recovery. The ONNX backend supports Qwen2.5 and Qwen3. The runtime improvements update and pin LiteRT-LM to 0.17.1.
Related tools
Related tools listed on this site.
LongCat
Model and API platform. This review reads the official LongCat-2.0 technical article and its linked API product page, covering code understanding/generation, reasoning and interaction purposes; the original chat UI was not reviewed in operation.
豆包工作
Doubao Work is Doubao’s productivity-focused AI product. Its primary official description presents goal-driven task decomposition and tool use for research, content creation, data processing and application building.
InfCode
Enterprise AI coding agent from TokensInfinity. Its site describes using enterprise code and knowledge through IDE, extension and CLI forms for completion, cross-file changes and engineering tasks.
提示工程指南
Prompt Engineering Guide is a learning resource for prompt design and LLM applications; its Chinese site covers techniques, examples, agents and limitations.
秒听
Tinrec is a meeting-notes and speech-to-text tool. Its public site describes live recordings, audio/video files and video-link transcription, followed by transcripts, meeting notes, speaker summaries and action items.
讯飞译制
iFlyrec Yizhi is a tool for audiovisual translation, AI dubbing and subtitle production, with timecode matching and output described on its official page.
Trancy
A reading and video assistant for language learning. Its site describes bilingual subtitles, selected-text and full-page translation, sentence analysis and personal learning materials.
通义听悟
The site calls it an AI assistant for work and study, built on the Qwen model and audio/video AI, for recording and reading audio/video content. Docs say individuals can use it free after phone sign-up; enterprises can integrate via API.
图改改
TuGaiGai is an online editor focused on text within images. Its official page describes changing, adding and removing image text, font recognition, image translation and image/PDF conversion.
多维视界
Duowei Shijie is an audiovisual analysis tool whose official site describes transcription, translation, summaries, interactive learning and content rewriting.
Winston AI
AI-content detection and plagiarism-checking tool whose site describes text scans, OCR input, sentence prediction maps and reports.
CleverSee
Alibaba Cloud’s enterprise AI search and question-answering platform, describing web retrieval, private multimodal data and configurable applications.
Sources
This page is written from the official sources below. Open them to check the original text.
- GitHub READMEhttps://github.com/dineshsoudagar/local-llms-on-androidFetched
Verified ·
This page is written by AI from the official sources above and checked automatically. This site has not tested it. Found a mistake? Tell us.