Segment Anything Model (SAM)
Promptable image segmentation foundation model plus SA-1B, 1.1B masks on 11M images, released under Apache 2.0.
Brought the foundation-model recipe to segmentation. Click/box/text prompts give zero-shot masks on unseen image types, and a model-in-the-loop data engine produced the then-largest segmentation dataset.
- Date
- Wednesday, 5 April 2023
- Lab
- Meta
- Kind
- open-weights
- Access
- open weights
Figures
| Measure | Value | Measured by |
|---|---|---|
| SA-1B dataset | 1.1B masks / ~11M images 400x more masks than any prior segmentation dataset | company |
Sources
This record was checked against its sources on 6 October 2026. How we check