엔비디아, 1조 파라미터 ‘네모트론4' 개발키로...개방형 모델 승부수

NVIDIA plans to develop a next-generation artificial intelligence (AI) model with at least one trillion parameters, marking a significant step towards open-source models. After previously struggling to achieve competitive edge with its own open-source 'Nemotron' series, NVIDIA's push into open-weight model development could reshape the competitive landscape for major clients.

According to multiple sources cited by the Korean IT publication Dive Informatics on November 11, NVIDIA is targeting the development of 'Nemotron 4,' which is expected to have a parameter count of at least one trillion, roughly double the 550 billion parameters of its 'Nemotron 3 Ultra.' However, the exact release date for this model remains uncertain.

The upcoming model is anticipated to be smaller than recent open-source models from China, such as Alibaba's 'Qwen 3.8 Max' (2.4 trillion parameters) and Chinese large-scale open-source AI 'Kimi K3' (2.8 trillion parameters). Nevertheless, NVIDIA aims to compensate for its relatively smaller model size by focusing on advanced compression techniques that allow for impressive performance despite fewer parameters.

NVIDIA's Vice President of Generative AI, Carrie Bryce, stated in an email statement that the company believes in the need for advanced open-source models that support safety, security, and innovation for all businesses and nations as they continue to evolve.

NVIDIA has been developing its own AI models to enhance its hardware design. However, NVIDIA's latest model, the 'Nemotron 3 Ultra,' has performed relatively poorly in global AI performance evaluation platforms such as Arena LM and Arith-ethical Analysis, ranking outside the top 40 among all models. Despite this, NVIDIA is actively advancing the open-source model ecosystem to expand its Graphical Processing Unit (GPU) market demand.

As open-source models improve and become more widely applicable, more companies are likely to develop and operate their own AI models, ultimately driving an increase in NVIDIA GPU sales.