[K-Tech] LG AI model EXAONE 4.0 ranks 11th globally in intelligence index

LG AI Researchs Exaone 40 Courtesy of LG AI Research — LG's Exaone 4.0/ Courtesy of LG

SEOUL, July 31 (AJP) - LG AI Research's artificial intelligence model EXAONE 4.0 secured 11th place globally in the Intelligence Index evaluation by AI performance analysis firm Artificial Analysis, according to results announced Wednesday. The model ranked first among Korean AI models and fourth among publicly available open-weight models.

The ranking was based on seven high-difficulty performance metrics including reasoning and knowledge capabilities, expert-level scientific problem solving, coding abilities, and mathematical problem resolution. EXAONE 4.0 achieved seventh place in coding and 10th place in mathematics, earning an overall score of 64 points alongside Anthropic's Claude 4 Opus.

The model achieved these rankings despite being significantly smaller than many competing frontier AI models from global companies. LG AI Research ranked eighth among AI model development companies in the evaluation, which measures models against tasks such as MMLU-Pro, GPQA Diamond, LiveCodeBench, and AIME 2024.

EXAONE 4.0, released on the Hugging Face platform on July 15 for research and educational purposes, reached 500,000 downloads within two weeks of its public release. The 32B parameter model has since exceeded 550,000 downloads, setting a record for Korean AI models in terms of fastest adoption rate.

Both EXAONE Deep, released in March as Korea's first reasoning AI model, and EXAONE 4.0 have been included in Epoch AI's Notable AI Models list. This designation is used by Stanford University's annual AI report for comparing national and corporate AI competitiveness globally.

Lee Hong-rak, co-research director of LG AI Research, presents at the LG AI Talk Concert 2025 on July 22/ Courtesy of LG

Candice Kim candicekim1121@ajupress.com