LAION-5B
Dataset
Image-Text
5.85 billion image-text pairs — the fuel of open image generation, and its scandal.
You're looking at a dataset record: a training-data collection — what's in it and which models learned from it.
LAION's web-scraped image-text pairs trained Stable Diffusion and much of open multimodality. A 2023 Stanford finding of CSAM links forced its withdrawal and the cleaned Re-LAION release — the case study in web-scale data's legal and ethical exposure.
How this connects to the rest of the stack
This record's links at a glance — click a group to jump to it.
Used to train
Related
Sources
Record reviewed as of 2026-01-15 — see Methodology.