Ollama
A local model runner for pulling and serving open-source LLMs on your own machine. It suits developers who want to download a model, serve it locally, and connect other tools to it.
Best local LLM apps: 9 self-hosted tools in one list
If you want to run a large language model on your own machine, these are the nine self-hosted apps in the AI Download Center catalog: Ollama, LM Studio, Jan, GPT4All, LocalAI, Open WebUI, AnythingLLM, Text Generation WebUI and PrivateGPT. Seven are open source, LM Studio is free, and Open WebUI is free and source-available, and each entry links to its official site so you can download it from the source.
This list isn't ranked, and we haven't benchmarked these apps. It groups them by the job they do, so you can find the one that fits how you work: some run models, some give you a chat interface, and some focus on chatting with your own documents.
Quick overview
| App | Type | Price type | Best fit |
|---|---|---|---|
| Ollama | Local app | Open source | Pulling and serving models on your own machine |
| LM Studio | Desktop app | Free | Downloading and testing models in a desktop UI |
| Jan | Desktop app | Open source | An open-source desktop app for chat and agents |
| GPT4All | Desktop app | Open source | Private, offline chat in a desktop app |
| LocalAI | Local server | Open source | An OpenAI-compatible API on your own hardware |
| Open WebUI | Self-hosted web app | Free (source available) | A browser-based chat front end for local models |
| AnythingLLM | Self-hosted web app | Open source | A workspace for chatting with your documents |
| Text Generation WebUI | Local web app | Open source | An extensible interface for experimenting with many models |
| PrivateGPT | Local app | Open source | Retrieval-augmented chat over private files |
Model runners and local APIs
LocalAI
An open-source local inference stack that exposes an OpenAI-compatible API. It suits developers whose code already talks to the OpenAI API and who want to point it at a server they run themselves.
Desktop apps
LM Studio
A desktop tool for downloading, running and testing local models offline. It suits people who would rather browse and try models in a desktop app than work in a terminal.
Jan
A local AI app for running LLMs, agents and productivity workflows without cloud dependencies. It suits people who want an open-source desktop app that keeps their work off the cloud.
GPT4All
An open-source local AI application from Nomic AI for offline models and private chat. It suits people who want a desktop chat app that runs offline, with conversations kept on their own machine.
Web interfaces you host yourself
Open WebUI
A feature-rich chat interface for self-hosted local model workflows and AI assistants. It suits people who already run models locally and want a chat interface they can open in a browser.
Text Generation WebUI
An extensible local web interface, by Oobabooga, for running many transformer-based language models. It suits tinkerers who want to try many different models and extend the interface.
Chat with your own documents
AnythingLLM
A self-hosted AI workspace from Mintplex Labs for chat, document ingestion and local knowledge-base workflows. It suits individuals or teams who want to load their own documents into a workspace and ask questions about them.
PrivateGPT
A local retrieval-augmented chat application for private, file-based AI work. It suits people who need to query sensitive files without sending them to a hosted service.
Open models to run in them
Each of these apps needs a model. The catalog lists these open-weight models, and each links to its maker's official page:
- Llama 4 (Meta): a natively multimodal mixture-of-experts open model family for local and private AI.
- Llama 3.2 (Meta): a small open model family (1B and 3B) suited to on-device and lightweight local inference.
- Qwen3 (Alibaba Cloud): a multilingual open model family for reasoning, chat, coding and agents.
- Gemma 4 (Google): a compact open model family with vision support across sizes, built for local and edge use.
- Phi-4 (Microsoft): a small model optimized for on-device and low-resource use.
- DeepSeek V4 (DeepSeek): an open mixture-of-experts model for reasoning and coding.
- Mistral Large 3 (Mistral AI): an open-weights frontier model released under Apache 2.0.
Which model works in which app, and what hardware it needs, depends on the app and the model. Check both official pages before you download.
Local coding assistants
If you're running models locally for code, the catalog also lists these open-source coding assistants:
- Aider: a terminal-based coding assistant for local workflows and code changes across repositories.
- OpenHands: a coding agent for autonomous task execution and local development automation.
- Cline: an autonomous coding agent for VS Code and JetBrains, also available as a CLI.
- Continue: a coding assistant for VS Code and JetBrains. It's no longer maintained after joining Cursor, but the final 2.0.0 release is still available under Apache 2.0.
How to choose
- Want a desktop app with no terminal? Start with LM Studio, Jan or GPT4All.
- Want to serve a model to other tools or your own code? Look at Ollama or LocalAI.
- Want a browser chat interface for models you already run? Look at Open WebUI or Text Generation WebUI.
- Want to ask questions about your own files? Look at AnythingLLM or PrivateGPT.
Many setups combine a runner with a separate interface. Before you pair two tools, check each one's official documentation to confirm they work together.
How we chose
Every tool on this page comes from the hand-picked part of the AI Download Center catalog: tools we add and check one by one, not a test lab. We haven't benchmarked or ranked these tools. The catalog also includes community model cards imported automatically from Hugging Face; those aren't covered by this guide and link to Hugging Face rather than to an official site. The descriptions, price types and links come from our catalog entries, and each link goes to the project's official site or repository. No listing on this page is sponsored.
See the full directory
Browse the self-hosted apps, open models and coding assistants in the catalog.