YOUR DEVICE. YOUR CONVERSATION.

Local AI Chat

A small language model, running right here in your browser.

Local inference

Choose your model

Load one model at a time. Smaller models need less memory; larger ones can handle more demanding questions.

Qwen2 · 0.5BSmallest download. A starting point for short, multilingual conversations.~1 GB GPU memoryLlama 3.2 · 1BCompact general assistant. Works without half-precision GPU support.~1.2 GB GPU memoryQwen2 · 1.5BMore capacity for writing and explanations. Needs more device memory.~1.9 GB GPU memory

Open this page in your browser to check device compatibility.

The first load downloads hundreds of MB from Hugging Face and the model runtime from MLC. Cached files may be reused on later visits. Loading can take several minutes.

Model details & license

Your messages are processed on this device. No account, API key or inference server is used.

Remember this chat on this deviceOff by default. Downloaded model files and saved chats are separate.

No model loaded

What would you like to explore?

Choose a model and load it to start. Ask a question, work on a draft, or think through a piece of code.

Enter to send · Shift+Enter for a new lineMessage space: 0%

Local models can make mistakes and do not browse the web. Verify important answers.