Parti

Training compute

5.1×10²³ FLOP

Parameters

20B

Published

Jun 22, 2022

Parti is an AI model developed by Google Research (United States), first published in June 2022. It works in the image generation domain, on tasks such as text-to-image and image generation.

Training it took an estimated 5.1×10²³ FLOP of compute (estimation method: operation counting). The model has 20,000,000,000 parameters. It was trained on roughly 4.7T datapoints. Training ran on Google TPU v4. The compute alone is estimated at $427K in 2023 dollars.

Access: Unreleased. Its weights are not openly released. The reference paper has 1,476 citations. Epoch AI rates the confidence of this record as confident.

Full record

Organization: Google Research
Country of organization: United States
Domain: Image generation
Task: Text-to-image, Image generation
Training compute: 5.1×10²³ FLOP
Compute estimation method: Operation counting
Parameters: 20,000,000,000
Dataset size: 4.7T
Training hardware: Google TPU v4
Training cost (2023 USD): $427K
Numerical format: BF16
Model accessibility: Unreleased
Open weights: No
Citations: 1,476
Epoch confidence: Confident