<ul><li>NVIDIA released NVILA, a family of open visual language models focusing on accuracy and efficiency.</li><li>NVILA outperforms GPT-4o Mini in video benchmark and shows competitive performance against other open models.</li><li>NVIDIA plans to release code and models for NVILA to enable reproducibility.</li><li>NVILA uses the 'scale then compress' technique to balance accuracy and efficiency in visual language models.</li></ul>

NVIDIA’s NVILA Beats GPT-4o Mini and Llama 3.2, Redefines Open Standards in Vision AI

Discover more