Try it: chat with a local AI
Start a free, no-signup conversation with an open-weights language model running live on our own machine.
Who do you want to chat with?
Pick one or both. You can talk to them together in a single conversation.
Conversation settingsThese shape how each bot speaks to you. You can change them per chat.
Run this chat model on your own computer
The live chat runs a small open model on one server, so it goes offline at times. Our step-by-step guide installs the same kind of model on your own machine with Ollama, on CPU alone. No GPU, no account, no subscription.
What you get from the guide
The guide gives you one-click install scripts for Windows, Mac and Linux. They install Ollama, pull an uncensored Qwen3.5 4B model, and check that it answers.
It also links a single HTML chat page you can open from your disk, with a name, avatar and personality for your bot. Once the model is downloaded, everything works with the internet off.
The guide was written around a laptop with 8 GB of RAM. It also covers what changes if you have a GPU.
Common questions
- Is this AI chat free?
- Yes. The live chat needs no account. Running it yourself is free too, since Ollama and the model are both free to download.
- Do I need a GPU to run it locally?
- No. The setup in the guide runs on CPU alone, with at least 4 GB of available RAM.
- Why is the chat sometimes offline?
- The live chat runs on a single machine rather than a hosted service, so it is not available around the clock. Running the model yourself removes that limit.
- Is a local chat private?
- Yes. Once the model is downloaded, nothing is sent anywhere. It runs as a local process on your machine.