NVIDIA ChatRTX
A free chatbot that runs on your own RTX GPU, fully offline.
What is NVIDIA ChatRTX?
ChatRTX is NVIDIA's free desktop app that turns your own RTX graphics card into a local AI chatbot, no internet connection required after setup. Point it at a folder of PDFs, notes, or photos and ask questions about them directly, since everything runs on your hardware instead of a company's server. It is slower to set up than a web chatbot, but nothing you type or upload ever leaves your machine.
Originally launched as Chat with RTX in 2024, NVIDIA folded the app into the broader ChatRTX brand as local, on-device AI became a bigger selling point for its GPUs. It uses TensorRT-LLM to run models like Mistral or Llama locally, plus retrieval over your own files, so it doubles as a private alternative to cloud tools like NotebookLM. The catch is the hardware requirement: you need a recent RTX card with enough VRAM to make it worth installing.
Key features
- Runs fully offline on local RTX hardware
- Chat with your own PDFs, docs, and notes
- Search photo libraries by describing what is in them
- Support for multiple local open-weight models
- No account or subscription required
- Voice input support
Pros
- Total privacy, nothing is sent to a server ever
- No subscription or usage limits once installed
- Genuinely useful for searching your own files without cloud tools
Cons
- Needs a specific NVIDIA RTX GPU, useless without one
- Setup and model downloads are clunky compared to a web app
- Answer quality trails well behind cloud frontier models
