YaART

Training compute

8.2×10¹⁹ FLOP

Parameters

2.3B

Published

Apr 8, 2024

YaART is an AI model developed by Yandex (Russia), first published in April 2024. It works in the image generation domain, on tasks such as text-to-image and image generation.

Training it took an estimated 8.2×10¹⁹ FLOP of compute (estimation method: operation counting). The model has 2,300,000,000 parameters. Training ran on 160 NVIDIA A100 SXM4 80 GB.

Access: API access. Its weights are not openly released. Epoch AI rates the confidence of this record as speculative.

Full record

Organization: Yandex
Country of organization: Russia
Domain: Image generation
Task: Text-to-image, Image generation
Training compute: 8.2×10¹⁹ FLOP
Compute estimation method: Operation counting
Parameters: 2,300,000,000
Training hardware: NVIDIA A100 SXM4 80 GB
Chips used: 160
Training power draw: 126.5 kW
Model accessibility: API access
Open weights: No
Epoch confidence: Speculative