LAION-5B
LAION-5B is the largest public image-text dataset, assembled by LAION from Common Crawl and filtered with CLIP, a model that measures how well an image and a piece of text go together. That filtering keeps pairs where the caption genuinely describes the image and drops the rest.
The dataset contains only the URLs pointing to images on their original hosts, along with the captions, so you download the images yourself. Some of those links have decayed over time, which means you will not recover the full set.
Because the material was gathered from the open web at scale, review content-safety considerations before use. The metadata is licensed under CC BY 4.0, but each image stays under whatever terms its original host set, so the licence on the dataset does not clear you to reuse the pictures.
Related sources
Conceptual Captions (CC3M / CC12M)
A cleaned image alt-text dataset for automatic image captioning, produced by Google Research. Alt-text pulled from the web is filtered and generalised (hypernymed), replacing specific names with broader categories. Available in 3.3M (CC3M) and 12M (CC12M) variants.
DataComp
A benchmark and dataset collection for training multimodal models. It provides a shared pool of image-text candidates and an evaluation framework so teams can test different data-curation strategies against a common yardstick.
OBELICS
A web-scale dataset of 141M multimodal English web documents that interleave text and images, containing 353M images and 115 billion tokens. Unlike image-caption pair datasets, it keeps whole documents with text and images in their original order. Used to train the IDEFICS models.
PD12M (Public Domain 12M)
An image-text dataset built only from materials marked with a Public Domain Mark or released under Creative Commons Zero (CC0). 12.4M image-caption pairs, with a 3.3M subset called PD3M, sized to match the Conceptual Captions datasets while staying copyright-clean for commercial use.