HyperCLOVA 204B

Training compute

2×10²³ FLOP

Parameters

204B

Published

Sep 10, 2021

HyperCLOVA 204B is an AI model developed by NAVER (South Korea), first published in September 2021. It works in the language domain, on tasks such as language modeling/generation. It counts among the frontier models: the systems trained with the most compute of their moment.

Training it took an estimated 2×10²³ FLOP of compute. The model has 204,000,000,000 parameters. It was trained on roughly 560B datapoints. Training ran on NVIDIA A100. The compute alone is estimated at $442K in 2023 dollars.

Access: Hosted access (no API). Its weights are not openly released. The reference paper has 92 citations. Epoch AI rates the confidence of this record as speculative.

Full record

Organization: NAVER
Country of organization: South Korea
Domain: Language
Task: Language modeling/generation
Training compute: 2×10²³ FLOP
Parameters: 204,000,000,000
Dataset size: 560B
Training hardware: NVIDIA A100
Training cost (2023 USD): $442K
Model accessibility: Hosted access (no API)
Open weights: No
Citations: 92
Epoch confidence: Speculative