ruGPT-3.5 13B

Training compute

1.1×10²³ FLOP

Parameters

13B

Published

Apr 24, 2023

ruGPT-3.5 13B is an AI model developed by Sber (Russia), first published in April 2023. It works in the language domain, on tasks such as chat and language modeling/generation.

Training it took an estimated 1.1×10²³ FLOP of compute (estimation method: operation counting,hardware). The model has 13,000,000,000 parameters. It was trained on roughly 300B datapoints. Training ran on 512 NVIDIA A100,NVIDIA Tesla V100 SXM2 for about 1.1K hours.

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

Full record

Organization: Sber
Country of organization: Russia
Domain: Language
Task: Chat, Language modeling/generation
Training compute: 1.1×10²³ FLOP
Compute estimation method: Operation counting, Hardware
Parameters: 13,000,000,000
Dataset size: 300B
Training hardware: NVIDIA A100, NVIDIA Tesla V100 SXM2
Chips used: 512
Training time: 1,080 h
Model accessibility: Open weights (unrestricted)
Open weights: Yes
Epoch confidence: Confident