JIJ Releases Major JijModeling 2 Updates to Streamline Mathematical Model Development and Debugging

2026/8/6

New Flat AST-based compiler architecture delivers compilation speed and memory efficiency matching or exceeding JijModeling 1.

TOKYO, Japan — JIJ Inc. (“JIJ”; headquarters: Minato-ku, Tokyo; CEO: Yu Yamashiro) announced the release of JijModeling 2 versions 2.6 and 2.7, available from July 30, 2026. The updates deliver substantial performance improvements for mathematical optimization practitioners and corporate R&D teams.

JijModeling 2 is JIJ’s next-generation mathematical optimization modeling tool. Compared with its predecessor, JijModeling 1, it features a fundamentally redesigned type system and greater expressive power, with a strong focus on usability in real-world applications.

The latest updates introduce a major overhaul of the compiler’s internal architecture. By adopting a Flat Abstract Syntax Tree (Flat AST), JijModeling 2 now provides compilation speed and memory efficiency comparable to or better than JijModeling 1, without compromising usability. JIJ has also redesigned the tool’s error messages and improved the workflow for updating mathematical models, further accelerating the overall model formulation process.

Background

Formulating mathematical models requires specialized knowledge, and identifying and resolving the causes of errors can consume a significant amount of development time.

At the same time, generative AI and large language models are increasingly being used to assist with mathematical model development based on natural-language requirements. As these integrations become more widespread, there is growing demand for modeling tools that can handle complex real-world data, process models more efficiently, and provide clearer information when errors occur.

To address these needs, JIJ developed JijModeling 2 by redesigning its syntax and APIs from the ground up to improve safety, readability, and usability. While the redesign made the tool more capable and convenient to use, it also introduced a new challenge: when working with large-scale models or supporting rapid iteration through an LLM, JijModeling 2 did not initially match the speed and memory efficiency of its predecessor.

In response, JIJ extensively adopted a Flat AST, an efficient internal representation that significantly improves compilation speed and memory efficiency. The company also made error messages easier to understand and introduced a comprehensive online Error Code Index. Together, these improvements enable users to move more quickly from model validation to production use.

About JijModeling 2

JijModeling 2 is a modeling tool for translating business and operational requirements into mathematical optimization models. It is designed primarily for data scientists working with mathematical optimization, corporate R&D teams, and existing JijZept users.

Building on JIJ’s experience in mathematical optimization software development, JijModeling 2 provides the following core capabilities to support the development and practical use of mathematical models.

Separation of Models and Parameters
A model can function as a schema, allowing users to modify its structure independently of the size of the instance data.

Solver-Independent Modeling
JijModeling 2 uses OMMX Message, a data exchange format for mathematical optimization. This allows users to switch freely between JIJ and third-party solvers and select the most appropriate solver for each problem.

Type Checking and Automatic Constraint Detection
JijModeling 2 includes a dedicated type system that helps detect modeling errors at an early stage. It also automatically identifies constraint patterns that can be processed more efficiently.
Intuitive Mathematical Representation
Users can interactively inspect their defined mathematical models as mathematical expressions in Jupyter Notebook or an integrated development environment.

Key Improvements in the Latest Updates

The latest releases preserve the advanced functionality of JijModeling 2 while introducing the following major improvements.

Significantly Improved Compilation Performance
By adopting a Flat AST *1 architecture, JIJ has improved compilation speed and memory efficiency to a level comparable to or better than JijModeling 1.
The improvement provides a more responsive modeling experience, including when working with large-scale mathematical models.
Clearer Error Messages and a New Error Code Index
JIJ has revised the error messages in JijModeling 2 to provide more specific information about the nature and cause of each error.
The company has also published a comprehensive online Error Code Index, which explains the likely causes of individual errors and how to resolve them. This reduces the effort required to diagnose and correct errors, enabling users to develop mathematical models more efficiently.

These improvements are also expected to accelerate model formulation when using JijZept AI *2, JIJ’s natural-language modeling assistant, or other large language models. By shortening the trial-and-error cycle, the updates support faster model validation and a smoother transition to practical use.

*1. Flat AST: A technique in which the Abstract Syntax Tree representing a program’s structure is flattened and laid out in memory to improve processing speed and reduce memory consumption.

*2. JijZept AI: JIJ’s AI capability for assisting with mathematical model development based on requirements entered by users in natural language.

Looking Ahead

JIJ will continue to improve the performance and usability of JijModeling 2 while strengthening its integration with generative AI technologies, including JijZept AI.

By leveraging its rapid development capabilities to deliver new value to users more quickly, JIJ aims to create an environment in which anyone can apply mathematical optimization to real-world business operations.

About JIJ Inc.

JIJ Inc. is a technology company that helps organizations address complex planning and optimization challenges across business and society, with mathematical optimization at its core while also leveraging AI and quantum technologies.

Across industries including manufacturing, transportation and logistics, telecommunications, energy, and materials development, JIJ formulates customer challenges as computable models and supports decision-making in areas such as planning, allocation, scheduling, and search.

Through the development and delivery of software products and open-source technologies for modeling, solving, development, and execution, JIJ is building an environment in which a broader range of companies and developers can apply advanced optimization technologies.

Media Contact

JIJ Inc. Communications Team
Email: pr@j-ij.com

Product Inquiries

Email: info@j-ij.com

JIJ Community on Discord: https://discord.gg/Km5dKF9JjG