Mitsubishi Electric Corporation has announced the development of a language model specifically designed for manufacturing processes that operate on edge devices. According to the company’s announcement, the AI technology, branded as Maisart®, has been pre-trained using data from Mitsubishi Electric’s own internal operations to support applications tailored to specific manufacturing domains. The firm claims that the model employs a unique data-augmentation method to optimise responses for user-specific industrial tasks.
This development reportedly addresses growing concerns around the significant computational and energy demands typically associated with large language models (LLMs), which have seen rapid uptake across various industries. Mitsubishi Electric emphasises that its new model is compact enough to run on hardware with constrained computing capabilities—such as edge devices used directly within factory environments—and can be deployed on-premises, which is important for managing data privacy and confidentiality, especially in sectors like call centres.
The language model builds on a publicly available Japanese base model, fine-tuned with proprietary data from Mitsubishi Electric’s factory automation business areas. This approach aligns with the company’s broader digital manufacturing strategies, including its work on the e-F@ctory platform that utilises edge computing to optimise real-time data processing on the factory floor. Mitsubishi Electric considers edge computing essential for timely and resilient responses within smart factory environments.
The company’s announcement fits into a wider context of Mitsubishi Electric’s AI and digital transformation efforts. Earlier in 2025, Mitsubishi Electric signed a memorandum of understanding with Amazon Web Services (AWS) to collaborate on cloud services, aiming to incorporate generative AI into digital platforms like Serendie, focusing on smart buildings and manufacturing solutions. This partnership seeks to reduce data centre carbon footprints and advance energy management systems, reflecting increased corporate emphasis on sustainability alongside digital innovation.
While Mitsubishi Electric’s Maisart AI has been behind commercialised products like MELSOFT VIXIO, an AI-powered visual inspection tool, and cutting-edge interfaces such as the ARIA pre-engineered work cell, its latest language model project appears to prioritise operational efficiency and data privacy in manufacturing domains. Independent industry analysis suggests that domain-specific AI models targeted at edge devices can help address latency and data security challenges posed by reliance on cloud-centric, large-scale AI systems.
In parallel with industrial AI innovations, Mitsubishi Electric’s research wing has been developing more energy-efficient AI methodologies, including techniques that reduce computational requirements and enhance model adaptation efficiency, contributing to greener AI implementations. This broader push towards “green AI” underscores the industrial sector’s efforts to marry technological advances with sustainability goals.
Nevertheless, the practical impact and uptake of such domain-specific language models within manufacturing ecosystems will depend on factors including integration with existing infrastructures, ease of use, and demonstrable gains in productivity and cost-efficiency. As generative AI continues to evolve rapidly, robust evaluation of proprietary models versus general-purpose alternatives will be crucial for end users in highly sensitive and data-critical environments.
In summary, Mitsubishi Electric’s new language model for edge devices represents a step in the company’s ongoing efforts to embed AI more deeply into manufacturing processes while addressing energy efficiency and data privacy concerns. This move complements the firm’s broader digital and cloud partnerships and reflects an industry-wide trend towards more customised, efficient AI applications tailored to specific operational contexts.
Source: Noah Wire Services
Subscribe to Industry Updates
