Chat with RTX

ChatRTX is a free Windows desktop demo from NVIDIA for consumers with RTX-powered PCs who want a personal AI chatbot that can answer questions using local files without sending data to the cloud.

Key capabilities include:

• Local retrieval-augmented generation (RAG) that searches selected folders and generates responses grounded in your content, with source attribution
• Support for querying local data such as .txt, .pdf, .docx, .xml, plus images and GIFs
• GPU-accelerated inference using NVIDIA TensorRT-LLM on GeForce RTX (30 Series or later) for fast, offline responses
• Compatibility with open-source LLMs such as Mistral 7B and Llama 2, plus mention of upcoming model support (including Llama 3 and Gemma)

Pricing: Free

What makes it notable: it runs fully on-device on RTX hardware, pairing RAG with TensorRT-LLM acceleration to keep personal data local while delivering low-latency answers compared with cloud-based chatbots.