Local AI with a visible boundary.

Once you have downloaded a model, chat runs entirely on your device. Chats, projects, attached files and models are stored locally, and everything you write or attach is encrypted there.

CuriousLM is designed for private everyday work, not only model demonstrations. It combines local generation with projects, document and photo understanding, thinking mode, sandboxed tools, alternate versions of replies, and export of everything you create. Replies show where the model was unsure and what it nearly said, so you can see its guesses and take a different path.

Read the complete product explanation for supported workflows, architecture, and device limitations.

Local inference does not mean the application never uses a network. The first visit, model downloads, updates and privacy-limited aggregate analytics use the internet.

CuriousLM is created and published by Ebenezer Don. Product pages and guides use named authorship so readers can distinguish maintained CuriousLM material from anonymous or automatically assembled pages.

The guide library is written to explain trade-offs, not to claim that local AI is always the right tool. Device limits, setup requirements, online exceptions, and uncertain behaviour are included where they affect a decision.

• Privacy and network actions
• Runtime and model licences
• Local AI glossary