ChatGPT alternatives for offline chat and documents

Open WebUI is for people who want a ChatGPT-style web interface they host themselves, alone or for a household or team. It runs on your computer or server through Docker, works offline with local models through Ollama, and can also connect to OpenAI and Anthropic models.

Beyond chat it handles document search, voice, images and tools, and administrators can set user roles and group permissions.

LibreChat replaces ChatGPT conversations with a shared workspace that switches between local backends and hosted APIs. Create specialized agents, work with files and code artifacts, and search earlier messages. Its authentication and shared agent controls suit a team server.

Cherry Studio covers everyday ChatGPT writing and file questions in a desktop app. Reusable assistants keep instructions for recurring work, and simultaneous conversations let you compare models. Connect local backends or hosted providers and add external tools through MCP.

Chatbox offers a familiar chat interface across desktop, mobile and the web. Its desktop knowledge base supports questions over saved documents, while Ollama connections let you use a local model instead of the ChatGPT service.

Jan is the closest thing to the ChatGPT app on your own computer. It is a free desktop chat app that downloads open models such as Llama, Gemma and Qwen and runs them offline, so you keep the familiar chat window without a subscription.

When you still want a hosted model, the same app connects to GPT, Claude, Gemini or Mistral, and custom assistants cover tasks you repeat.

Msty Studio is a desktop workspace for people who want local and hosted models side by side. You can compare answers in split chats, keep reusable prompts and personas, and attach document collections called Knowledge Stacks to a conversation.

Local models run through MLX, llama.cpp or Ollama. Picking a hosted model sends that chat to its provider.

AnythingLLM fits people who mostly used ChatGPT to ask questions about their own files. Drop PDFs, text and Word documents into a workspace and get answers with source citations. The desktop app needs no account.

It works with local backends such as Ollama and LM Studio and adds agents and memory. A Docker version serves several users from one server.

big-AGI suits questions where you want more than one model's answer. Beam generates independent responses, and Merge combines them using your instructions; split-pane chats and conversation branches let you examine different approaches.

LM Studio suits people who want to find, download and chat with local models in one desktop app on macOS, Windows or Linux. Downloaded models run offline, and you can ask questions about files on your computer, the way you would attach a document to a chat.

It also serves models through a local OpenAI-compatible API, so other apps on your machine can use them. The desktop app has its own license terms rather than an open-source license.

5ire puts document questions and reusable prompts in a desktop chat workspace. Variable-based prompts help with repeated tasks, and MCP connections let a model use files, databases or other tools you configure.

Newelle brings document search and voice conversations into a Linux desktop chat app. It can work with files or run commands from chat, with permissions that limit what the assistant can read or change.

Off Grid AI suits people who want model chat and document questions on a phone as well as a computer. Its project knowledge base creates document embeddings on the device, while compatible models can use saved-document search and other tools.

RecurseChat lets you import ChatGPT history and continue those conversations with a chosen local model. A floating chat window stays beside your work, with file questions backed by on-device document search.

NativeMind connects Chrome to Ollama for conversations about several open tabs at once. It also accepts PDFs, images and screenshots, so you can compare sources and ask file questions alongside the pages you are reading.

Brave Leo answers questions about the page or document you are reading in the browser, including PDFs, Google Docs and Google Sheets. Connect a chosen model for those conversations, with Brave Search available when an answer needs current information.

Alpaca lets you download Ollama models and use several models within one conversation. Ask questions about documents and images, create custom personas, or bring a webpage or YouTube transcript into the chat.

Locally AI covers ChatGPT on the phone. It runs language and vision models on recent iPhones and iPads, and on Macs, with on-device voice conversations and no account.

After a model download it works offline, and Siri and Shortcuts can call it. It is available from the App Store and the Mac App Store.

Maid gives Android users a chat app that runs downloaded GGUF models on the phone. You can adjust each conversation's model settings and export chats; document research is outside the workflow described here.

Data Formulator covers ChatGPT-style table analysis in a dedicated chart workspace. Load spreadsheets or database tables, refine charts directly and branch an investigation without losing earlier results.

DB-GPT suits ChatGPT users analyzing business data across databases and spreadsheets. Its agents generate SQL and Python, calculate metrics and produce charts or reports; you connect the data sources and model backend.

Enchanted covers everyday ChatGPT-style conversations through an Apple app with voice prompts, image attachments and spoken replies. It connects to your Ollama server, which handles the model processing.