According to Foresight News, NVIDIA (NVDA.O) CEO Jensen Huang announced the launch of a complete series of models based on Llama, utilizing the NVIDIA Llama Nemo Tron language foundational model, named Nemotron. Additionally, NVIDIA introduced NVIDIA Cosmos, a foundational model designed to understand the real world, utilizing the NVIDIA Llama Nemo Tron language foundational model.
The Cosmos model can accept text, images, or video prompts to generate virtual world states, providing video output tailored to the unique needs of autonomous driving and robotics applications. Developers can leverage Cosmos to generate AI feedback for reinforcement learning, thereby improving policy models and testing performance in various scenarios. NVIDIA is collaborating with the global developer ecosystem to drive the next wave of advancements in physical AI. Cosmos is the world's first foundational model, trained on 20 million hours of video, with a focus on physical dynamics and human interactions.