#Japan #Tokyo #ABEJA #GENIAC #LLM Development

ABEJA's Strategic Move into Next-Gen AI Systems

In a groundbreaking development, ABEJA, headquartered in the Minato Ward of Tokyo, has announced that its proposal for the 'Long-Context Compatible Base Model and AI Agent Development Research' has been accepted by the NEDO (New Energy and Industrial Technology Development Organization). This initiative is part of a larger project called the 'Post-5G Information Communication System Infrastructure Strengthening Research and Development Project / Competitive Generative AI Base Model (GENIAC)'.

ABEJA has committed to implementing artificial intelligence to foster a 'rich world', focusing on mission-critical services where accuracy and reliability are paramount. The company aims to leverage insights gained from previous project phases to enhance the utilization of long language models (LLMs) across various applications within enterprises. This project represents ABEJA's third consecutive participation in GENIAC, confirming its commitment to contributing to the implementation of LLMs into society.

Background of ABEJA

Since its founding in 2012, ABEJA has focused on pushing the boundaries of AI research and development, particularly in deep learning and quantum computing. The company's leadership emphasizes the importance of developing robust digital platforms to support mission-critical operations for commercial enterprises. Their flagship platform, the ABEJA Platform, stands at the forefront of this mission, providing enterprises with the necessary tools for AI integration.

With the explosive growth of generative AI, businesses globally are exploring ways to harness its potential. ABEJA's engagement in this GENIAC initiative forms part of the greater strategy by Japan's Ministry of Economy, Trade and Industry, aiming to bolster the country's capabilities in foundational model development.

Goals and Objectives

For this third phase, ABEJA's objective is straightforward yet ambitious. They are set to develop LLMs that not only achieve high precision and usability but are also designed for mission-critical operations. As many enterprises remain cautious about adopting LLMs due to concerns about data handling, the new models will prioritize security and effectiveness. ABEJA plans to collaborate with enterprise clients to implement these LLMs and AI agents in real-life critical tasks, proving their reliability and efficiency.

The project involves creating a local working environment for LLMs with fewer than tens of billions of parameters—enabling faster processing and improved security. Notably, ABEJA will enhance the long-context performance of its models, allowing them to manage extensive and complex information demands, such as those arising from long-term operational histories and multiple manuals.

Enhancing AI Capabilities

Another primary focus is improving the planning abilities of AI agents, equipping them to break down and sequentially accomplish complicated tasks. ABEJA aims to implement tools that can autonomously select and execute critical processes, enhancing overall workflow efficiency. The company has already recognized the benefits of AI agents since the first phase of their involvement, embedding a unique approach into the ABEJA Platform. In this third phase, the AI agents will see significant enhancements in their capabilities.

In terms of performance metrics, ABEJA is targeting benchmarks that would place their models above established giants like OpenAI's GPT-4o and Alibaba's Qwen3, particularly in specific evaluation criteria for LLMs published in Japan.

Significance of the Third Phase

As the project progresses from August 2025 to February 2026, ABEJA will continue to collaborate with domestic enterprises to apply these newly developed LLMs and AI agents effectively. This venture promises not only to optimize operations in critical business segments but also to validate the potential of AI through empirical results within enterprise environments.

A commitment to safety, precision, and reliability makes ABEJA's mission reflective of the company's core philosophy: 'Implementing a rich world through partnership between humans and AI'. Their digital platform initiatives illustrate how technology can transform industry structures and operational efficiencies whenever executed carefully and responsibly.

For more information about ABEJA and their initiatives, visit their homepage at ABEJA Inc. or check the NEDO's official announcement for the latest updates on GENIAC.