Windows
Windows 11, 64-bit. Installs for your user account, with no administrator rights.
Download.exe installer
Local AI · Windows and Linux · free An
indoor AI.
Chat, dictate, listen, proofread, learn: ChatonLM puts what you do with AI into a single app. It lives on your computer, not in the cloud.
Version 0.9 beta · free and open source · no account
Six tools. One window.
Take a look around: every tool is one click away, and they all share your models.
Chat. Chat with your models and drop in images, PDFs or code. Reasoning models get their own Think button.
ModelHub. The catalog of open models. ChatonLM shows you the ones that fit your graphics card and memory, and installs them in one click.
Dictation. Press a shortcut and speak, wherever you are. The text is transcribed on your computer and copied, ready to paste.
Voice. Have answers read aloud in a natural voice, or talk hands-free: you speak, the model answers out loud, then listens again.
Proofread. Correct or rephrase a text in the app. Elsewhere, a shortcut corrects your selection and copies the result, ready to paste.
Study. Pick a topic: your model writes a course, with lessons, examples and quizzes, based on your own documents if you like.
Six apps elsewhere. One here.
Usually you need one app to talk to a model, another to dictate, a third to have a text read aloud, a fourth to proofread it. Each has its own settings, its own models to download all over again, sometimes its own account. ChatonLM installs once, and everything works together.
Six installs, six sets of settings, six updatesOne app, one set of settings
The price
No
subscription.
Assistant, dictation, voice, proofreader, study tool: elsewhere, each one has its own subscription, and the bill soon comes to about sixty euros a month*. ChatonLM is free. No limited trial, no “Pro” version, no credit card.
- €0per month
- 0accounts to create
- 0ads or trackers
* Total of the public prices of five common online services (assistant, dictation, read aloud, proofreader, study tool), billed yearly, as checked in October 2026.
And that’s not all.
Six more reasons. Follow the thread down the page.
Your voice stays home
Dictation and voice mode are transcribed by your own computer. Elsewhere, your voice is often sent off to a server.
Nothing to set up
Install the app, open it: no server to start, no container. The engine that runs the models is included. If it’s already on your machine, ChatonLM uses it, with the models you’ve already downloaded.
It knows your machine
ModelHub reads your graphics card and memory, then points you to the models that fit your machine. No more 20 GB downloads for nothing.
Linux doesn’t wait its turn
The same tools on Linux and on Windows. Many AI apps stop at Windows and macOS.
The code is open
ChatonLM is open-source software under the MIT license. Anyone can read what the app does, change it, and check that it hides nothing.
It speaks your language
Interface in English, French and Spanish, natural voices in all three languages, dictation in 99. ChatonLM is designed in France.
What leaves the house: almost nothing.
No account, no telemetry, no ads. And when ChatonLM does go online, you know why.
Stays on your computer
- Your conversations and attachments
- Your voice: dictation, read aloud and voice mode
- The texts you proofread and your courses
- Your settings and history
Goes over the internet
- The catalog and downloads of models and voices
- Checking for updates, all of them signed
- Web search, optional, with an online account
Ready to adopt it?
Choose your system. Installation takes a minute, and the model engine is included.
The first public version, the 0.9 beta, is on its way. Write to us and we’ll let you know when it’s out.
Linux · AppImage
Most recent distributions. A single file, no installation. Updates download on their own and install when you choose.
Download.AppImage file
Linux · .deb
Ubuntu 22.04, Debian 12 and their derivatives. Installs like any system package.
Download.deb package
Version · 64-bit PC (x86-64). A graphics card speeds up the models but isn’t required. No macOS version for now. On Linux, global shortcuts need KDE Plasma 6 or GNOME 48; elsewhere, set them in your desktop’s settings.
Common questions
What is ChatonLM?
ChatonLM is a free, open-source local AI app for Windows 11 and Linux. It brings six tools together in one window: chat with a language model (LLM), a model catalog, voice dictation, voice (read aloud and voice mode), proofreading and study. Models run on your own computer, with no account and no subscription.
Is it really free?
Yes. ChatonLM is open-source software under the MIT license: no paid version, no subscription, no ads. The open models you download are free too.
Do I need a graphics card?
No. Models also run on the processor, just more slowly. With a graphics card they run faster and can be bigger; ModelHub tells you which ones fit your machine.
Does it work without an internet connection?
Yes, once your models, voices and speech recognition are downloaded. Chat, dictation, voice, proofreading and study all work offline. The internet is only used to browse the catalog, to download models and voices, to check for updates and, if you turn it on, for web search.
Which models can I use?
Models you can download and run locally, such as Gemma, Qwen, Mistral or Llama. ModelHub lists them in one catalog, tells you which ones fit your machine, and installs them in one click.
Does ChatonLM work with Ollama?
Yes. ChatonLM is a desktop app for Ollama: the engine it includes is Ollama, unmodified. If Ollama is already installed on your computer, ChatonLM connects to it and picks up the models you’ve already downloaded. ChatonLM is an independent project: it is not affiliated with Ollama, nor endorsed by it.
Is it as capable as the big online AIs?
Not at everything. A model that fits on your computer is smaller than the ones behind the big online services. For chatting, summarizing, proofreading, translating, studying or dictating, it does the job very well, and your data stays with you.
What about macOS?
Not for now. ChatonLM is available for Windows 11 and Linux, 64-bit.
What else do I need to install?
Nothing to get started. The engine that runs the models is included; you just pick a model in ModelHub, and one click is enough. Voices and speech recognition are downloaded the first time you use them.
Windows shows a warning during installation
ChatonLM is a young app: its installer doesn’t yet have a certificate that Windows recognizes, hence the “Windows protected your PC” screen. Click “More info”, then “Run anyway”. The code is open, and every update is signed.