Country
South Korea
Tracked models
92
Frontier models
2
Largest training run
3.9 x 10^24 FLOP
92 tracked AI models have at least one contributing organization based in South Korea. 2 of them are frontier models — among the largest training runs of their moment. The largest is EXAONE 4.5, at 3.9 x 10^24 FLOP (April 2026). The most recent tracked release is from June 2026.
It hosts 12 tracked GPU clusters.
Models (80+)
EXAONE 4.53.9 x 10^24 FLOPEXAONE 4.0 (32B)2.7 x 10^24 FLOPEXAONE 1.01.7 x 10^24 FLOPK-EXAONE1.5 x 10^24 FLOPSolar Open 100B1.4 x 10^24 FLOPEXAONE 3.5-R 32B1.3 x 10^24 FLOPEXAONE Deep 32B1.3 x 10^24 FLOPEXAONE 3.5 32B1.3 x 10^24 FLOPMi:dm 200B1.2 x 10^24 FLOPSolar Open2 250B1.1 x 10^24 FLOPEXAONE 3.5-R 7.8B4.3 x 10^23 FLOPEXAONE Deep 7.8B4.2 x 10^23 FLOPEXAONE 3.5 7.8B4.2 x 10^23 FLOPEXAONE 3.04 x 10^23 FLOPTri-21B2.9 x 10^23 FLOPHyperCLOVA 204B2 x 10^23 FLOPKonan LLM 41B1.7 x 10^23 FLOPHyperCLOVA 82B1.5 x 10^23 FLOPVARCO LLM 2.0 base1.2 x 10^23 FLOPEXAONE 3.5-R 2.4B9.5 x 10^22 FLOPEXAONE Deep 2.4B9.4 x 10^22 FLOPEXAONE 3.5 2.4B9.4 x 10^22 FLOPTrillion-7B9.3 x 10^22 FLOPEXAONE 4.0 (1.2B)8.7 x 10^22 FLOPProtT5-XXL7.4 x 10^22 FLOPVARCO LLM 2.0 small Finetuning6.7 x 10^22 FLOPVolcano 13B4.6 x 10^22 FLOPProtBERT-BFD3.9 x 10^22 FLOPGigaGAN3.9 x 10^22 FLOPKonan LLM 13B3.9 x 10^22 FLOPProtT5-XXL-BFD3.7 x 10^22 FLOPVARCO LLM KO/EN-13B-IST ver.12.7 x 10^22 FLOPProtT5-XL-U501.9 x 10^22 FLOPProtBERT-UniRef7.3 x 10^21 FLOPProstT53.1 x 10^21 FLOPTiny Recursive Model (TRM-Att)3.1 x 10^20 FLOPRQ-Transformer (3.8B params ImageNet dataset)2.9 x 10^20 FLOPDiscriminator Guidance2.2 x 10^20 FLOPRoseTTAFold All-Atom (RFAA)2.1 x 10^20 FLOPFusion in Encoder1.3 x 10^20 FLOPRQ-Transformer (1.4B params ImageNet dataset)1.1 x 10^20 FLOPscHyena8.6 x 10^18 FLOPCLUE3.5 x 10^18 FLOPRQ-Transformer (LSUN-cat dataset)2.9 x 10^18 FLOPDensePhrases2.1 x 10^18 FLOPLMRec2 x 10^18 FLOPA.X (Adot) 18B1.4 x 10^18 FLOPBITTERS7.8 x 10^17 FLOPEI-REHN-1000D1.1 x 10^16 FLOPPyramidNet2.3 x 10^15 FLOPVAETKIDecember 2025A.X K1December 2025HyperCLOVA X SEED 32B ThinkDecember 2025EXAONE Path 2.0July 2025Gauss2November 2024SK Telecom TelcoNovember 2024FvFoldNovember 2024LongVUOctober 2024SO3LROctober 2024Protein-MambaSeptember 2024RoseTTAFold2-LiteSeptember 2024Solar ProSeptember 2024LUXIA-21.4B-AlignmentMay 2024Zephyr 141B-A39BApril 2024ProteinINRJanuary 2024Solar-10.7B (Solar Mini)December 2023Mi:dm 7BNovember 2023Samsung Gauss LanguageNovember 2023Samsung Gauss CodeNovember 2023Samsung Gauss ImageNovember 2023HyperCLOVA XAugust 2023A.X (Adot) 7BAugust 2023RFdiffusionJuly 2023EXAONE 2.0July 2023LLaMA-13B (LoRA finetuned)May 2023LLaMA-7B (LoRA finetuned)May 2023LLaMA-65B (LoRA finetuned)May 2023LLaMA-33B (LoRA finetuned)May 2023A.X (Adot) 39BMay 2023DiT-XL/2 + Discriminator GuidanceNovember 2022
GPU clusters (12)
Samsung SSC42.5K H100eNHN Cloud's National AI Data Center1K H100eNAVER Corp Sejong706 H100eSamsung SSC-21555 H100eNaver DGX Superpod353 H100eSK Telecom Titan Phase 2328 H100eGcore data center Phase 1320 H100eKT SuperPOD250 H100eKT Internal MI250 Cluster232 H100eSamsung SSC-21 Scalable Module45.4 H100eAlibaba Cloud in South Korea Phase 2Alibaba Cloud in South Korea Phase 1
SourceEpoch AI, 'AI Models'. Published online at epoch.ai. Retrieved 2026-07-29 from https://epoch.ai/data/ai-models. Licensed under CC BY 4.0.
Other countries