PaLI-X

Training compute

5.6×10²³ FLOP

Parameters

55B

Published

May 29, 2023

PaLI-X is an AI model developed by Google Research (United States), first published in May 2023. It works in the multimodal, language, vision and video domain, on tasks such as image captioning, video description, character recognition (ocr) and visual question answering.

Training it took an estimated 5.6×10²³ FLOP of compute. The model has 55,000,000,000 parameters.

Access: Unreleased. Its weights are not openly released. It is built on top of UL2. The reference paper has 282 citations. Epoch AI rates the confidence of this record as likely.

Full record

Organization: Google Research
Country of organization: United States
Domain: Multimodal, Language, Vision, Video
Task: Image captioning, Video description, Character recognition (OCR), Visual question answering
Training compute: 5.6×10²³ FLOP
Parameters: 55,000,000,000
Model accessibility: Unreleased
Open weights: No
Base model: UL2
Citations: 282
Epoch confidence: Likely