waypoints.si The sky Every system The lead Countries Forecast revisions
Masked Autoencoders ViT-H
Meta, 11 Nov 2021.
- Lab
- Meta
- Released
- 11 Nov 2021
- Kind
- Vision, Semantic segmentation
- Training compute
- 4.6 × 10²⁰ FLOP
- Parameters
- 632.0 million
- Organisations
- Facebook AI Research
- Country
- United States
Built on
- ViT-Huge/14, 2020
See it in the 3D sky Paper or announcement
Data: Epoch AI (Creative Commons Attribution), releases too new for it from OpenRouter, and METR. Updated 25 Sep 2026.