Imagen 3 technical report
Google publishes the Imagen 3 report on a latent diffusion model that its human evaluations rank above other leading text-to-image models.
Documents a latent diffusion text-to-image model with quality and responsibility evaluations covering safety and representation. Authors state it was preferred over other leading models at evaluation time. 260+ contributors listed.
- Date
- Tuesday, 13 August 2024
- Lab
- Google DeepMind
- Kind
- paper
- Access
- closed API
Date is arXiv v1; product availability (Gemini, Vertex AI, ImageFX) was staged around the same period and is not itemised here.
Sources
This record was checked against its sources on 6 October 2026. How we check