NVIDIA NVLM-1

NVIDIA NVLM 1.0 is a family of frontier-class open multimodal large language models developed by NVIDIA to achieve state-of-the-art performance across both vision-language and text-only tasks. By coupling a high-capacity large language model backbone with a robust vision encoder, the architecture excels at jointly processing complex visual inputs—such as high-resolution photographs, detailed charts, diagrams, tables, and handwritten mathematical notes—alongside natural language prompts. Uniquely, training the model on multimodal datasets also enhances its core text-only performance, yielding improvements in multi-step reasoning, mathematical problem-solving, and coding tasks. By open-sourcing the model weights and training codes, NVIDIA provides researchers and developers with a powerful, competitive alternative to leading proprietary multimodal systems for advanced AI experimentation and deployment.