The original purpose of this tutorial is to ensure offline privacy and portability on your cell phone by using DeepSeek-R1 Models that are good or not are benevolent. Vomiting for 10 days... Nope... .10 minutes to bring you a tutorial on how to install DeepSeek-R1 locally on your cell phone.
Cell phone use DeepSeek-R1 Three programs
Prior to installation, the reader is provided with a popularization of the common scheme for using the DeepSeek-R1 model on cell phones:
Option 1, online use:downloading DeepSeek Official APP, or download a third-party APP that supports the DeepSeek-R1 model, for example:mock ingot (burnt as offering in worship),Nano AI Search, Cyberstarfire,Tiangong AI... Wait, it needs to be networked for use.
Option 2, AI Chat Tool Access API:Not for newcomers, e.g. from the DeepSeek official website ,Silicon-based flow ,volcano engine, after applying for the API, access to native AI chat casing tools available in the phone, for example:BotGem,Chatbox,GPT Mobile,NextChat(Web page deployment),Page Assist(Browser plugin)... And so on.
Option 3, cell phone local deployment:It is this teaching note, while not limited to cell phones to install DeepSeek-R1, many other models can be installed.
Cell Phone Installation DeepSeek-R1 Description
Difference between the end-side model that comes with the phone and the self-deployment model
In fact, cell phone installation of large models is not a very new topic, for example, focusing on the side of the end of the small size model of the ModelBest, or Ali's MNN-LLM-Android: MNN Project, can be installed in the phone in the big model, but they are too specialized, has been in the experimental groping stage, with the beginning of the year 2024 many vendors will be end-side model commercialization, cell phone local installation of the big model of the application is also more. For example, the recent hot news:Cell phone AI ushered in the era of "intelligent body": Samsung S25 joins hands with Smart Spectrum to open a new era of audio and video calls, which then utilizes an end-side model customized for mobile devices.
But a model developed specifically for the mobile side of the house will be different in its area of specialization than a larger model deployed locally by the phone itself:
End-side model customized for mobile devicesIt emphasizes the concept of "mobile assistant" and is good at executing specific tasks designed by the cell phone manufacturer, while executing rich types of tasks.
Self-installed large models locally on the phoneGenerally it is a "compressed version" of a really big model, an "enhanced version" of a certain aspect of the ability, in general, "general knowledge" ability is strong, but can not perform the regular automated operation of the phone, such as Naturally, he can't open his cell phone's address book.
Notes on installing large models locally on cell phones
Installation of equipment:IOS/Android
Storage space:2~4G
Applications:Copywriting, paragraphs, role-playing, quick summaries, translations
Supports local installation of DeepSeek-R1 model app on cell phones
Pocket AI--Android
Pocket AI (PocketPal AI Chinese version) is a powerful offline AI assistant designed to allow users to talk to AI anytime, anywhere. The project is based on Small Language Models (SLMs) and runs on cell phones without an internet connection, especially adapted to Chinese user experience. Pocket AI supports a variety of small language models, including DeepSeek, Danube, etc., with intelligent memory management, real-time performance display and other features. The project currently supports Android devices, and the iOS version is under development.
PocketPal AI--IOS/androidYou need to log in to the official app store
PocketPal AI is an open source mobile app designed to bring Small Language Models (SLMs) directly to your phone for both iOS and Android users. It provides a web-independent AI chat experience that ensures user privacy and supports automatic memory management and a rich selection of models. Developed by Asghar Ghorbani, this program integrates with the Hugging Face model library, allowing users to download and use a wide range of models directly through the app for different needs in chatting and problem solving.
Open Intelligence is a company dedicated to providing open source AI solutions, and its main product, Apollo, allows users to interact with their private AI backend directly through their cell phones. The platform not only supports individual users to autonomously manage their AI backends, but also provides support for a variety of AI application scenarios, such as chatting, data processing, etc. Apollo is designed to simplify users' interactions with AI and to improve the efficiency of AI applications in daily life and work.Open Intelligence seeks to provide users with a more flexible, private AI experience through its open architecture. private AI experience.
Distilled versions of the Llama 3.1, Qwen, and DeepSeek R1 series in multiple sizes are available.
Fullmoon is an application designed for iOS devices and aims to provide the ability to chat privately with native large language models. The app is optimized for Apple Silicon and is supported on iPhone, iPad and Mac. Users' chats are saved locally and the appearance of the app can be customized.Fullmoon leverages Apple's MLX Swift framework, an array framework for machine learning research on Apple Silicon. The app supports multiple language models, including Llama 3.2 1B, Llama 3.2 3B and DeepSeek-R1-Distill-Qwen-1.5B-4bit.
Enchanted is an open source iOS, macOS and visionOS app designed for multimodal chat with private models such as Llama 2, Mistral, Vicuna and Starling. The app is compatible with Ollama and provides a secure, private and multimodal experience that users can use on all devices in the iOS ecosystem.Enchanted's goal is to provide a product that allows users to have unfiltered, multimodal interactions in a private environment.
Tifa-Deepsex-14b-CoT--Android APK Direct Download
Tifa-Deepsex-14b-CoT is a large model based on Deepseek-R1-14B deep optimization, focusing on role-playing, fictional text generation, and Chain of Thought (CoT) reasoning ability. Through multi-stage training and optimization, the model solves the original model's problems of insufficient coherence in long text generation and weak role-playing ability, which is especially suitable for creative scenarios that require long-range contextual correlation. By fusing high-quality datasets and incremental pre-training, the model significantly enhances the contextual association ability, reduces non-answers, and eliminates Chinese-English mixing and increases domain-specific vocabulary for better performance in role-playing and novel generation. In addition, the model supports 128k ultra-long contexts for scenarios requiring deep dialog and complex authoring.
This is a version of Deepseek-R1-14B that is deeply optimized for long-form fiction and roleplaying scenarios, and has a simple Android client available for download.
Mobile Local Installation Other Model Software
Ai2 OLMoE--IOS
Ai2 OLMoE is an open source iOS app developed by the Allen Institute for AI (Ai2, Allen Institute for Artificial Intelligence) to provide AI models that run entirely on the device. The app utilizes Ai2's open-source OLMoE models, which are able to run offline without a cloud connection, ensuring user privacy and data security.Ai2 OLMoE is not only suitable for researchers exploring how to improve local AI models, but also provides developers with tools to rapidly prototype new AI experiences.
Llamao is a private and offline running Llama AI chatbot designed to provide users with an intelligent assistant service without internet connection. Unlike ChatGPT, Llamao runs entirely on the user's device, ensuring absolute privacy and security of user data. Whether it's writing, brainstorming or problem solving, Llamao helps users accomplish their tasks efficiently while avoiding the risk of data leakage.Llamao requires no subscription fees, so users are free to utilize its powerful AI features and enjoy the convenience of having no network restrictions.
English 
简体中文 
日本語 
Deutsch 
Português do Brasil