Large Generative AI Development Support Program

ABCI has been implementing the “ABCI Grand Challenge” program to support the challenges of the most important issues in the field of artificial intelligence, which are only possible through enormous computing power, and important social issues utilizing artificial intelligence, big data, and quantum computation. In FY2023, the “Large-scale Language Model Building Support” program was conducted against the backdrop of the growing demand for large-scale language models. In FY2024, we will expand the scope of the program from language models to include all aspects of building the “Large-scale Generative AI Models,” including large-scale language models and diffusion models.
ABCI limits the number of parallel jobs and their execution time in order to avoid excessively long waiting times for jobs. However, this limitation makes it difficult to run jobs that require a huge amount of computation, such as large-scale generative AI models. This program enables users who wish to build large-scale generative AI models to reserve a larger number of compute nodes. Please note that ABCI points should be obtained in charge by the ABCI users to execute this program. Unlike the Grand Challenge, no free ABCI points will be granted.

Public Application Guidelines


Resources to be provided



Proposal Review

Proposals will be reviewed by the Large-Scale Generation AI Research and Development Support Program Review Committee, consisting of experts, through document review and discussion based on the results of the review.

The main items to be reviewed are as follows:

Application Deadlines

How to apply

Acceptance of applications has been closed. Thank you for so many applications.

Please submit your application using this form ( (Sorry, Japanese only.)

Please refer to the application form below for input items. (Sorry, Japanese only.)

Contact information
ABCI Contact:

Accepted Proposals (Sorry, Japanese only.)

Organization Proposal Number of nodes Days
国立研究開発法人産業技術総合研究所 継続事前学習に基づく汎用性および有用性の高い日本語大規模言語モデルの構築 36 140
国立研究開発法人情報通信研究機構 世界最大級の高品質日本語データ構築法の探索およびそのデータに適した事前学習ハイパーパラメータの探索 32 100
国立研究開発法人産業技術総合研究所 日本語音声と話し言葉に強い音声音響基盤モデルの構築 20 70
株式会社ELYZA 実践的応用を視野に入れた日本語大規模言語モデルの開発: 知識と系列長の拡張 40 70
国立研究開発法人産業技術総合研究所 データの倫理問題を根本解決するテキストからの画像生成モデル構築 40 21
理化学研究所 AI for Science実現に向けた科学基盤モデルの構築 2 140

Listed in order of receipt of applications.

Events Archive