EXAONE 4.0 (32B)

Training compute

2.7×10²⁴ FLOP

Parameters

32B

Published

Jul 15, 2025

EXAONE 4.0 (32B) is an AI model developed by LG AI Research (South Korea), first published in July 2025. It works in the language domain, on tasks such as language modeling/generation, question answering, code generation and 2 more.

Training it took an estimated 2.7×10²⁴ FLOP of compute (estimation method: reported). The model has 32,000,000,000 parameters. It was trained on roughly 14T datapoints. Training ran on 512 NVIDIA H200 SXM for about 3.2K hours.

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

Full record

Organization: LG AI Research
Country of organization: South Korea
Domain: Language
Task: Language modeling/generation, Question answering, Code generation, Quantitative reasoning, Translation
Training compute: 2.7×10²⁴ FLOP
Compute estimation method: Reported
Parameters: 32,000,000,000
Dataset size: 14T
Training hardware: NVIDIA H200 SXM
Chips used: 512
Training time: 3,192 h
Training power draw: 701.4 kW
Numerical format: FP8
Model accessibility: Open weights (non-commercial)
Open weights: Yes
Epoch confidence: Confident