ENCORD LAUNCHES WORLD'S LARGEST OPEN SOURCE MULTIMODAL DATASET TO ACCELERATE MULTIMODAL AI DEVELOPMENT
Major New Resource Drives Innovative Approach to Model Training to Democratize Multimodal AI Development, Dramatically Reduce Training Time and Compute Requirements for Builders
SAN FRANCISCO, Oct. 17, 2025 /PRNewswire/ -- Encord, the data infrastructure company for physical and multimodal AI, today announced that it has released the largest open source multimodal dataset in the world, empowering AI developers everywhere – from the largest labs to the youngest startups – to build and deploy sophisticated multimodal AI systems. The dataset has helped Encord develop a revolutionary methodology – called EBind – that will enable developers to train multimodal AI models on a single GPU within a matter of hours rather than the days such training processes currently require.
Together, Encord's multimodal dataset and EBind methodology represent a significant leap forward for democratizing multimodal AI while reducing time-to-market and increasing competitiveness for AI builders everywhere.
Encord Co-Founder and CEO Eric Landau said, "Multimodal AI is the next major leap for our industry, with the power to teach robots, self-driving cars, drones and other systems to recognize and make inferences from their physical environments using the same combination of senses that humans use. Until now, however, frontier research in multimodal AI has been the domain of large, well-resourced AI labs and enterprises, with smaller teams struggling to keep up. The multimodal dataset and hyper-efficient EBind training methodology we've announced today will vastly reduce the time and compute power needed to develop, train and deploy multimodal AI systems – and will help to unleash the next wave of innovation in this space."
Encord's EBind methodology utilizes carefully curated, high-quality data and relies on a single encoder per data modality (images, video, audio, text, etc.). This innovative approach offers a new paradigm for training multimodal AI models, driven by data quality rather than raw compute power. Encord's research has demonstrated that its EBind methodology enabled a simple 1.8 billion parameter multimodal model to outperform models up to 17 times larger – with training completed within hours on a single GPU.
Encord's multimodal dataset itself consists of 1 billion data pairs and 100 million data groups across 5 modalities, including text, image, video, audio and 3d point clouds.
Encord Co-Founder and President Ulrik Stig Hansen said, "While compute power gets the headlines, the real battlefield in multimodal AI is data. Our new multimodal dataset and EBind methodology show that, in the next phase of multimodal AI's evolution, winning organizations will be those that adopt innovative approaches to data curation and dataset construction – not just those that throw escalating levels of compute power at the problem. Encord is pleased to play a lead role in demonstrating the power of high-quality, curated data in advancing multimodal AI with our new dataset – and by making it open source for the entire AI community."
Captur AI CEO Charlotte Bax said, "The open source release of Encord's new multimodal dataset is an exciting step for our industry, and for multimodal AI builders everywhere. For Captur AI – with our unique, on-device approach to photo validation for enterprises – the dataset opens new possibilities for improving performance on image quality measures for our shared models across various verticals. We're always looking at ways to augment datasets for our on-device models to achieve better handling of edge cases, and Encord's new dataset offers a powerful pathway to accomplish that goal."
Pre-registration for Encord's new multimodal dataset can be accessed here.
About Encord
Encord is a multimodal AI data infrastructure platform that enables AI labs, human data companies and enterprise AI teams to curate, label and manage data for AI models and systems at scale, using agentic and human-in-the-loop workflows to enable these teams to work seamlessly with every type of data. Encord works with customers including Synthesia, Toyota, Zipline, AXA Financial, Northwell Health and more.
Media Contact:
Chris Clemens
Nexios Communications Strategies
chris@nexioscomms.com
View original content to download multimedia:https://www.prnewswire.com/news-releases/encord-launches-worlds-largest-open-source-multimodal-dataset-to-accelerate-multimodal-ai-development-302586944.html
SOURCE Encord