LG AI Model 'Exaone 4.0' Ranks 11th Globally, 4th Among Open Models

Intelligence Index Evaluation by Artificial Analysis
Ranked First Among Korean Models
Competing with World-Leading AI Models
Seventh in Coding, Tenth in Mathematics, Eleventh Overall
Surpassed 500,000 Downloads in Just Two Weeks
"Proven to Possess Intelligence and Potential"

LG AI Research Institute announced that its self-developed artificial intelligence (AI) model, 'Exaone 4.0,' ranked 11th globally in the Intelligence Index evaluation by Artificial Analysis, a global AI performance analysis organization. Among Korean models, it ranked first, and among open-weight (publicly available) models, it ranked fourth.

Choi Junggyu, Head of AI Agent Group at LG AI Research, is presenting at the 'LG AI Talk Concert 2025' held on the 22nd at the Convergence Hall of LG Science Park in Magok, Gangseo-gu, Seoul. Photo by LG

Artificial Analysis calculates the Intelligence Index by comprehensively evaluating seven high-difficulty performance indicators, including reasoning and knowledge capabilities, expert-level scientific problem-solving skills, coding ability, and mathematical problem-solving skills.

Exaone 4.0 ranked seventh in coding and tenth in mathematics, placing within the top ten in each category. In the overall ranking, it scored in the 64-point range alongside 'Claude 4 Opus,' the top model from Anthropic in the United States, securing 11th place. LG AI Research Institute also ranked eighth among AI model development companies.

Notably, Exaone 4.0 drew attention by ranking fourth in the overall ranking of open-weight models. This is significant as it stands shoulder to shoulder with global frontier AI models whose sizes are several to several dozen times larger.

Amid positive global market feedback, Exaone 4.0's 32B version (with 3.2 billion parameters) surpassed 500,000 downloads within just two weeks of its release. It currently stands at 550,000 downloads, marking the shortest period among domestic AI models.

Exaone 4.0 was released as an open-weight model on July 15 through Hugging Face, a global open-source AI platform, for research, academic, and educational purposes. It is Korea's first hybrid AI model that combines general and reasoning models into one. Previously, Exaone 4.0, together with 'Exaone Deep,' Korea's first reasoning AI model released by LG AI Research Institute, was also listed in the 'Notable AI Models' list selected by Epoch AI, a nonprofit AI research organization in the United States. This list is used as a reference for comparing national and corporate AI competitiveness in the annual AI report published by Stanford University in the United States.

Lee Honglak, co-head of LG AI Research Institute, emphasized, "This result proves that LG's Exaone has the capability and potential to compete with the world's top frontier AI models," adding, "We will continue to challenge ourselves to develop world-class AI models to achieve our goal of becoming one of the top three global AI powerhouses."