Do Son 
NVIDIA has officially unveiled Project G-Assist — an experimental AI assistant designed for users of GeForce RTX graphics cards. This innovation is now accessible through the NVIDIA App and operates on PCs equipped with RTX 30, 40, and 50 series GPUs, requiring a minimum of 12 GB of video memory. Support for laptops is anticipated in future updates.
G-Assist functions as a local AI assistant that facilitates the management of system settings, optimizes gaming performance, monitors temperature and component loads, and adjusts peripheral devices — all through voice or text commands. Unlike cloud-based solutions, G-Assist operates entirely offline, relying on the GPU’s resources without requiring an internet connection.
The project is powered by a compact language module, trained on Llama and refined for executing commands locally. When activated using the Alt+G shortcut, the assistant momentarily harnesses a portion of the GPU’s power to process the request before returning it to its primary task — such as running a game.
Beyond its core capabilities, G-Assist can export performance metrics, suggest overclocking settings, identify system bottlenecks, and interact with devices from Logitech G, Corsair, MSI, and Nanoleaf — offering control over lighting, fan speed, and more.
Of particular interest is the plugin module. NVIDIA has released tools on GitHub for crafting custom extensions to G-Assist. With simple JSON configurations and Python scripting, enthusiasts can build additional features — from voice-controlled Spotify playback and Twitch stream monitoring to integration with Google Gemini for advanced dialogue and information retrieval.
The company emphasizes that G-Assist is part of the broader RTX AI initiative, which also underpins the NVIDIA ACE technology. Third-party developers such as MSI, HP, and Logitech are already creating their own AI assistants based on this framework.
Integration with low-code platforms like CrewAI and LangFlow further expands its potential, allowing users to design automated workflows and user interfaces without any programming knowledge.
Project G-Assist is available now via the Home → Discovery tab in the NVIDIA App. Currently, it supports only the English language and desktop GeForce RTX GPUs. NVIDIA has committed to continuously enriching the assistant’s capabilities by incorporating user feedback and regularly introducing new commands and features.
