Country
Japan
Tracked models
60
Frontier models
1
Largest training run
1.2 x 10^24 FLOP
60 tracked AI models have at least one contributing organization based in Japan. 1 of them is frontier models — among the largest training runs of their moment. The largest is PLaMo-100B, at 1.2 x 10^24 FLOP (June 2024). The most recent tracked release is from September 2025.
It hosts 25 tracked GPU clusters.
Models (60)
PLaMo-100B1.2 x 10^24 FLOPnekomata-14b2.6 x 10^23 FLOPPLaMo-13B1.2 x 10^23 FLOPCalm2-7B5.5 x 10^22 FLOPtsuzumi 7B upgrade 20244.2 x 10^22 FLOPWeblab-10B3.6 x 10^22 FLOPFugaku-LLM3 x 10^22 FLOPLUKE1.8 x 10^22 FLOPStockmark-13B1.7 x 10^22 FLOPbilingual-gpt-neox-4b1.2 x 10^22 FLOPBERT-RBP1.4 x 10^20 FLOPOpenCALM9 x 10^19 FLOPMothra3.7 x 10^19 FLOPB2T connection (16L)2.8 x 10^19 FLOPSubformer (122M)5.1 x 10^18 FLOPSubformer (96M)4 x 10^18 FLOPSubformer (83M)3.5 x 10^18 FLOP(ensemble): AWD-LSTM-DOC (fin) × 5 (WT2)6.7 x 10^17 FLOPADP-FAIRSEQ + NGRAMRES1.5 x 10^17 FLOPAWD-LSTM-DOC (fin) (23M)4.3 x 10^16 FLOPAWD-LSTM+WT+Cache+IOG (WT2)3.2 x 10^15 FLOPNeocognitron2.7 x 10^8 FLOPCognitron5.2 x 10^6 FLOPSignal Processing TransformerSeptember 2025Llama-Primus-Nemotron-70BApril 2025SHIFT-SUVApril 2025IEV2MOLOctober 2024TakaneSeptember 2024pKALMSeptember 2024PLTNUMSeptember 2024ProtT3May 2024SwallowApril 2024NEC cotomiApril 2024HAM-TTSMarch 2024Karakuri LMJanuary 2024ProthyenaJanuary 2024Qarasu-14BDecember 2023ElyzaDecember 2023tsuzumi 7BDecember 2023CTM (CIFAR-10)October 2023Japanese-LM-3.6BAugust 2023PolySphere-1June 2023ADP-FAIRSEQ+NGRAMRESOctober 2022GPT-NeoX-JapaneseJuly 2022PFPMay 2022GraphBPApril 2022Japanese-GPT-1BJanuary 2022Japanese dialog transformersNovember 2021RNN + char4-MS-vecJuly 2019RNN + char3-MS-vecJuly 2019RNN + char2-MS-vecJune 2019(ensemble): AWD-LSTM-DOC (fin) × 5 (PTB)August 2018AWD-LSTM-DOC (fin) (37M)August 2018Spectrally Normalized GANFebruary 2018AWD-LSTM+WT+Cache+IOG (PTB)September 2017Bidirectional RNNNovember 1997Time-delay neural networksMarch 1989Hierarchical CognitronApril 1984Self-Organizing Nets of Threshold ElementsNovember 1972Decision tree adalineMay 1969
GPU clusters (25)
AIST ABCI 3.06.1K H100eAIST ABCI-Q2K H100eSoftBank CHIE-22K H100eSoftBank CHIE-32K H100eSakura's H100s Phase 12K H100eJCAHPC Miyabi1.1K H100eTSUBAME4.0960 H100eSoftbank SuperPOD646 H100eAIST ABCI 2.0303 H100eNEC Corp Japan Supercomputer293 H100eAIST ABCI 1.0275 H100eJapanese Research Institute 2 Supercomputer 1197 H100eUniversity of Tokyo Wisteria/BDEC-01 (Aquarius)114 H100eOsaka University SQUID106 H100eJAMSTEC ZettaScaler-2.2 Gyoukou81.05 H100eJapan Atomic Energy Agency and Quantum and Radiological Science and Technology HPE SGI860068.72 H100ePreferred Networks MN-264.68 H100eNagoya University Flow Type II (Furo)55.84 H100eJapanese Research Institute 1 Supercomputer 152.55 H100eSAKURA Internet 2019 V100 Supercomputer48.51 H100e"Center for Advanced Intelligence Project, RIKEN, RAIDEN"26.28 H100eGSIC TSUBAME 3.023.12 H100eUniversity of Tsukuba Cygnus20.21 H100eJAMSTEC ZettaScaler-2.0 Gyoukou12.97 H100eGSIC TSUBAME 2.58.4 H100e
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