News
EveryLLM

Korean

HyperCLOVA X, EXAONE. 13 models on EveryLLM so far.

Korean models are built for Korean, written in the Hangul alphabet grouped into syllable blocks, with word endings that change with politeness and grammar. South Korea's internet companies and conglomerates, led by Naver, LG, SK Telecom, KT and Kakao, along with start-ups such as Upstage, train their own models, and the government now funds a national programme.

Naver and LG

Naver announced HyperCLOVA in 2021 at 204 billion parameters, trained heavily on Korean, and launched HyperCLOVA X in August 2023 as the engine of its search and Clova services. LG AI Research, founded in 2020, builds the EXAONE series: EXAONE 3.0 (August 2024), a 7.8-billion-parameter model trained on 8 trillion tokens, was released openly for research and followed by reasoning and larger versions.

A national model

On 4 August 2025 the Ministry of Science and ICT chose five teams, led by Naver Cloud, SK Telecom, LG AI Research, NC AI and Upstage, from 15 applicants to build sovereign foundation models, with government support for GPUs and data. The teams are narrowed in stages as the project goes on. Supporters cite language and security; sceptics ask whether Korea can match the spending of the largest US and Chinese labs.

Still to be written: Clova (virtual assistant).

Further reading

The weekly digest

Every Monday: the model of the day, new pages on EveryLLM, and the conferences and anniversaries of the week ahead.