{"id":37123,"date":"2025-12-23T06:49:48","date_gmt":"2025-12-23T06:49:48","guid":{"rendered":"https:\/\/metaverseplanet.net\/blog\/?p=37123"},"modified":"2026-01-05T07:59:57","modified_gmt":"2026-01-05T07:59:57","slug":"game-playing-ai-nvidia-nitrogen-introduced","status":"publish","type":"post","link":"https:\/\/metaverseplanet.net\/blog\/game-playing-ai-nvidia-nitrogen-introduced\/","title":{"rendered":"Game-Playing AI: Nvidia NitroGen Introduced"},"content":{"rendered":"\n

Nvidia<\/strong> has introduced its new open-source AI model<\/strong>, NitroGen<\/strong>. Trained on over 1,000 games<\/strong>, the model can convert game footage directly into action. Here are the details…<\/p>\n\n\n\n

Nvidia<\/strong> has announced NitroGen<\/strong>, a new model that brings artificial intelligence<\/strong> to the gaming world. Developed as open source<\/strong>, this model not only guides players but can also play games directly on its own. NitroGen<\/strong> is based on the G-Assist<\/strong> approach previously introduced by Nvidia and will support over 1,000 games<\/strong>.<\/p>\n\n\n\n

AI Control for Over 1,000 Games<\/a><\/em><\/strong> At the core of NitroGen<\/strong> lies a “vision-to-action<\/strong>” approach that converts visual data directly into control inputs. The model was trained using over 40,000 hours<\/strong> of publicly available game videos, covering more than 1,000 games<\/strong>. During the training process, gamepad layouts<\/strong> displayed on-screen by streamers were analyzed to isolate joystick movements<\/strong> and button presses<\/strong>. Thus, the artificial intelligence<\/strong> learned which actions were being performed from in-game images.<\/p>\n\n\n\n

\"\"<\/figure>\n\n\n\n

On the technical side, NitroGen<\/strong> relies on a policy model based on GROOT<\/strong>, which Nvidia<\/strong> developed for its robotics<\/strong> work. However, instead of robot motor commands, it generates direct gamepad inputs<\/strong>. Nvidia<\/strong> states that this allows the model to perceive in-game situations and provide human-like reactions<\/strong>. In tests conducted, NitroGen<\/strong> offers up to a 52% increase in success<\/strong> compared to models trained from scratch, even in games it has never encountered before.<\/p>\n\n\n\n

NitroGen<\/strong>‘s scope is currently limited mainly to games played with a controller<\/strong>. Since the model is designed to predict gamepad actions<\/strong>, it is not directly used in keyboard-mouse oriented games. However, Nvidia<\/strong> points out that these limits can be expanded by developers thanks to its open-source<\/strong> structure.<\/p>\n\n\n\n

\"\"<\/figure>\n\n\n\n

NitroGen<\/strong> is also intended to be used in robotics<\/a><\/em><\/strong> and simulation<\/strong> fields outside of games. Reflexes<\/strong> and decision-making mechanisms<\/strong> learned from different game genres serve as a valuable data source for systems operating in the physical world. Nvidia<\/strong> sharing the entire model and datasets<\/strong> as open source<\/strong> could accelerate research in this area.<\/p>\n\n\n\n

You Might Also Like;<\/h3>\n\n\n