BEIT-3

Training compute

7×10¹⁹ FLOP

Parameters

1.9B

Published

Aug 22, 2022

BEIT-3 is an AI model developed by Microsoft (United States), first published in August 2022. It works in the multimodal, vision and language domain, on tasks such as object detection, semantic segmentation, image classification and 3 more.

Training it took an estimated 7×10¹⁹ FLOP of compute (estimation method: operation counting). The model has 1,900,000,000 parameters.

Access: Open weights (unrestricted). Its weights are openly available. The reference paper has 721 citations. Epoch AI rates the confidence of this record as likely.

Full record

Organization: Microsoft
Country of organization: United States
Domain: Multimodal, Vision, Language
Task: Object detection, Semantic segmentation, Image classification, Visual question answering, Image captioning, Language generation
Training compute: 7×10¹⁹ FLOP
Compute estimation method: Operation counting
Parameters: 1,900,000,000
Model accessibility: Open weights (unrestricted)
Open weights: Yes
Citations: 721
Epoch confidence: Likely