NVIDIA updates Project G-Assist with lighter model and 6GB RTX support

Project G-Assist receives a lighter on-device Artificial Intelligence model that cuts VRAM use by 40% and broadens support to RTX GPUs with 6GB or more.

At Gamescom NVIDIA delivered the first major update to Project G-Assist, its experimental on-device assistant for tuning RTX systems by voice or text. The headline change is a new, more efficient model that uses 40% less VRAM while retaining response accuracy. That efficiency translates into broader compatibility: G-Assist can now run on all RTX GPUs with 6GB or more of VRAM, including many laptops. The update also sharpens tool-calling intelligence and introduces a new G-Assist Plug-In Hub powered by mod.io to make discovering and installing extensions easier.

G-Assist acts as a central command center for PC tuning, designed to cut through buried control panels and utility menus. Users can run diagnostics to optimize game performance, display or chart frame rates, latency and GPU temperatures, and adjust GPU or peripheral settings such as keyboard lighting. Getting started requires the NVIDIA app and the Game Ready Driver issued on Aug. 19; after downloading the G-Assist update from the app home screen, press ´Alt+G´ to activate. NVIDIA said a follow-up update in September will add laptop-specific commands for features like NVIDIA BatteryBoost and Battery OPS.

The new G-Assist Plug-In Hub is a key part of the release. Built in collaboration with mod.io, the hub lets users ask G-Assist what plug-ins are available and install them using natural language thanks to a mod.io plug-in. NVIDIA showcased a G-Assist Plug-In Hackathon with finalists that include Omniplay for in-game lore and notes, Launchpad for app groups, and Flux NIM Microservice for generating on-device images. Plug-ins are based on JSON and Python, and NVIDIA´s Plug-In Builder simplifies creation by enabling natural language driven development. Hackathon winners were scheduled to be announced on Aug. 20.

NVIDIA also highlighted updates to RTX Remix, its modding platform for modernizing classic games. RTX Remix has grown to hundreds of active projects and more than 100 released mods with over 2 million downloads. At Gamescom NVIDIA named contest winners such as Painkiller RTX Remix, and previewed a path-traced particle system coming in September that brings fully simulated physics, dynamic shadows and realistic reflections to visual effects. That particle support will open particle creation to over 165 RTX Remix-compatible titles, and modders are already using linked workflows like RTX Remix plus ComfyUI to restore and remake textures with physically based materials and immersive path-traced lighting.

55

Impact Score

USPTO launches artificial intelligence pilot for pre-examination prior art search

The USPTO’s Artificial Intelligence Search Automated Pilot will provide applicants with automated prior art insights before substantive examination. The agency will test its internal artificial intelligence system’s ability to generate an Artificial Intelligence-Assisted Search Results Notice and guide early filing strategies.

Stitch Fix pilots generative artificial intelligence style experience

Stitch Fix launched a beta style visualization tool, Stitch Fix Vision, that uses generative Artificial Intelligence, proprietary data, and human stylists to render shoppers in shoppable outfits across real-world settings. The experience aims to scale personalization and streamline discovery beyond traditional virtual try-on.

IBM artificial intelligence news roundup

IBM’s artificial intelligence category page aggregates recent product launches, research, and partnerships spanning software, infrastructure, and governance. Highlights include TechXchange 2025 announcements, new enterprise models, and deployments across sports, education, and healthcare.

Generative artificial intelligence on Network World

Network World’s Generative Artificial Intelligence hub curates the latest news, features, and resources at the intersection of artificial intelligence and enterprise networking. Highlights span optical networking reliability, chip manufacturing moves, edge hardware, and cloud-based network operations.

Contact Us

Got questions? Use the form to contact us.

Contact Form

Clicking next sends a verification code to your email. After verifying, you can enter your message.