September-October. 2025 |

. AI, a leading Chinese AI company formerly known as Zhipu, is emerging as a serious contender in the global AI landscape.

This comes after the recent introduction of its latest large language model( LLM), GLM-4.5. The newly released model is gaining attention for achieving state-of-the-art( SOTA) performance, surpassing other open-source models.

Z. ai ' s first open-source release is built on a Mixture of Experts( MoE) architecture. The model comes in two versions: a 355 billion parameter flagship model and a 106 billion parameter GLM-4.5-Air.

With MoE architecture, the

model can scale up to 355 billion parameters without activating all of them simultaneously. Meanwhile, its agent-native design enables the model to autonomously plan complex tasks, generate data visualisation, and manage end-toend workflows.

The GLM-4.5 ' s performance in reasoning, coding, and agentic abilities was evaluated on 12 representative benchmarks. Its average score placed it third worldwide and first among domestic and opensource models.

In fact, the GLM-4.5-Air outperformed other models at the 100 billion parameter scale. In addition to its impressive generation speed, the GLM-4.5 series is a

KT CORPORATION has recently announced a partnership with Gyeonggi Province to develop a generative artificial intelligence( AI) platform for public administration. The project, valued at US $ 9.5 million, marks a first for the South Korean local government.

KT will oversee the platform ' s end-to-end development, which spans approximately 11 months. This involves configuring its Korean-specialised large language model( LLM), ' Mi: dm 2.0,' to meet the administrative functions of Gyeonggi Province.

The telecommunications company will also launch KT AI Studio, an integrated LLM operating platform, along with AI-powered services designed for public sector use. It aims to enhance operational efficiency by automating routine tasks such as document prepara- cost-effective solution, especially for large-scale deployments.

Its application programming interface( API) usage is priced at US $ 0.11 per million input tokens and US $ 0.28 per million output tokens.

" We are setting a new benchmark with GLM-4.5, demonstrating that cutting-edge performance can be open, efficient, and affordable," said Z. ai Chief Executive Officer( CEO) Peng Zhang.

" The first-principles approach to measuring artificial general intelligence( AGI) is to integrate more general intelligent capabilities without losing existing ones. GLM- 4.5 is our first complete realisation of this concept."

tion, meeting scheduling, and data retrieval.

KT is also working with five tech companies to support the platform ' s development. MT Data, Wisenut, Konan Technology, and Daishin Information & Communications( DSIC) are among them.

At the same time, the company is advancing its public sector AI initiatives through the Ministry of Science and Information and Communications Technology( ICT)' s ' Proprietary AI Foundation Model ' project.

Moving forward, future AI services will be introduced across four key sectors: law, safety, education, and healthcare. This includes AI-powered legal consultations, predictive policing, AI tutors, and administrative support for healthcare.

The project, according to KT

APPLE is set to launch ' Apple Manufacturing Academy ' in Detroit on Aug 19, 2025. This academy is a key part of its US $ 500 billion investment pledge in the United States( U. S.) over the next four years.

The initiative underscores Apple ' s longstanding commitment to fostering innovation and enhancing advanced manufacturing nationwide.

The academy provides free, expert-led programmes developed in partnership with Michigan State University. These programmes are aimed at industry players and small and medium-sized enterprises( SMEs).

" Apple works with suppliers in all 50 states because we know advanced manufacturing is vital to American innovation and leadership, stated Apple Chief Operating Officer( COO) Sabih Khan.

" With this programming, we ' re thrilled to help even more businesses implement smart manufacturing so they can unlock amazing opportunities for their companies and our country."

Training will be provided through on-site workshops in Detroit. These sessions will focus on machine learning and deep learning in manufacturing, as well as automation technologies, data analytics, and other advanced smart manufacturing practices.

Later this year, the programme will introduce online courses to improve accessibility. The virtual curriculum will emphasise practical skills such as project management and manufacturing process optimisation.

Sabih Khan

Public Sector Division Executive Vice President Yong-Gyu Yoo, represents the first full-scale adoption of generative AI by a local government. " It ' s about integrating data and offering real, practical support for digital innovation in public administration."

" KT will continue to expand AI services tailored to the practical needs of public institutions, centred on ' Mi: dm 2.0 ', and become a trusted partner leading the AI transformation in Korea," added Yoo. @ ND