neutral
Recently
Encord opens world’s largest open‑source multimodal dataset and debuts EBind single‑GPU training method
Encord’s new open dataset plus EBind allows large multimodal models on one GPU, democratizing AI training.
Encord officially released what it calls the world’s largest open‑source multimodal dataset, covering text, image, video, audio and 3D point cloud modalities. Alongside, it launched EBind a training methodology enabling a 1.8 billion parameter multimodal model to be trained in hours on a single GPU, claiming performance comparable to models 4–17× larger. The move aims to democratize advanced multimodal AI, making it accessible to smaller teams and startups.