NVIDIA has announced the groundbreaking NVIDIA NitroGen AI model, set to revolutionize the gaming world. Developed in collaboration with giants like Stanford and Caltech, this open-source model demonstrates incredible capabilities by independently playing over 1000 games.

How Does NVIDIA NitroGen AI Work?

Described by NVIDIA Research Director Jim Fan as “GPT for Action,” NitroGen is fundamentally based on translating visual data into action. This revolutionary model was trained using over 40,000 hours of publicly available game streams and videos. The goal was to teach the model to learn the player’s real-time interactions with their gamepad. This allowed the AI ​​to learn not only what to do, but also how to do it.

1/ NitroGen: NVIDIA's new image-to-action model!

NitroGen, a vision-action foundation model for generalist gaming agents that is trained on 40,000 hours of gameplay videos across more than 1,000 games.

Gaming is a significant factor in AI training. Google DeepMind trained AI… pic.twitter.com/Rk6io0WQaw

— Chubby♨️ (@kimmonismus) December 21, 2025

The technology is based on the GROOT N1.5 architecture, originally designed for robotics. This suggests that NitroGen could potentially undertake significant tasks not only in the virtual world but also in the physical world in the future.

NitroGen has been tested across a wide range of genres, from RPGs and platform games to battle royale and racing games. According to research, the model achieved a 52% higher success rate compared to custom models trained from scratch, even in games it had never encountered before. This demonstrates how well the model learns general game mechanics and “player instincts.”

NitroGen’s capabilities aren’t limited to games. This experience gained in thousands of games with different physics and mechanics holds great potential in robotics. For example, what is learned from this model could play a critical role in how a robot can make flexible and accurate decisions in an environment it has never encountered before.

Furthermore, NVIDIA has shared all the project’s outputs as open source. Model weights, datasets, and code are available to researchers and developers worldwide. This allows anyone to integrate NitroGen into their own projects or take it even further.

So, what are your thoughts on this new artificial intelligence technology from NVIDIA? Share your opinions in the comments!