ruDalle: Kandinsky 3.0

Training compute

2×10²⁰ FLOP

Parameters

11.9B

Published

Dec 11, 2023

ruDalle: Kandinsky 3.0 is an AI model developed by Sber (Russia), first published in December 2023. It works in the image generation domain, on tasks such as text-to-image and image generation.

Training it took an estimated 2×10²⁰ FLOP of compute (estimation method: operation counting). The model has 11,900,000,000 parameters. Training ran on NVIDIA A100.

Access: Open weights (unrestricted). Its weights are openly available. Epoch AI rates the confidence of this record as speculative.

Full record

Organization: Sber
Country of organization: Russia
Domain: Image generation
Task: Text-to-image, Image generation
Training compute: 2×10²⁰ FLOP
Compute estimation method: Operation counting
Parameters: 11,900,000,000
Training hardware: NVIDIA A100
Model accessibility: Open weights (unrestricted)
Open weights: Yes
Epoch confidence: Speculative