Beyond transformers: Nvidia’s MambaVision aims to unlock faster, cheaper enterprise computer vision

Join our daily and weekly newsletters for the latest updates and exclusive content on industry-leading AI coverage. Learn More Transformer-based large language models (LLMs) are the foundation of the modern generative AI landscape. Transformers aren’t the only way to do gen AI, though. Over the course of the last year, Mamba, an approach that uses Structured State Space Models (SSM), has also picked up adoption as an alternative approach from multiple vendors, including AI21 and AI silicon giant Nvidia. Nvidia first discussed the concept of Mamba-powered models in 2024 when it initially released the MambaVision research and some early models. This week, Nvidia is expanding on its initial effort with a series of updated MambaVision models available on Hugging Face. MambaVision, as the name implies, is a Mamba-based model family for computer vision and image recognition tasks. The promise of MambaVision for enterprise is that it could improve the efficiency and accuracy of vision ope ... Read full article.