AI industry ‘game changer’ CLEVI develops ‘Ivy’, a platform based on ultra-large-scale generative AI
Clevi Corporation (Clevi) announced that it would launch the open beta service for ‘Ivy’, its own ultra-large-scale generative AI developed using LLM (large language model) technology.
The company has officially launched its own ultra-large-scale generative AI model, ‘Ivy’, using LLM (large language model) technology. As a high-performance, low-cost model, Ivy delivers outstanding effectiveness and performance, with reasoning, specialised agent capabilities and a high instruction-following rate. Until now, South Korea had failed to achieve notable results in the AI market amid the dominance of OpenAI in the United States and DeepSeek in China. However, the official launch of ‘Ivy’ by the South Korean start-up CLEVI has demonstrated that South Korea can also develop AI using its own technological capabilities.
Ivy provides powerful capabilities, including △real-time learning and adaptation △multi-data processing △a code interpreter △expert model generation and △an agent support system, making it suitable for use across a wide range of fields, from individual users to enterprises.
CEO Lee Hwan-ho had been interested in the AI field and continued research and development even before founding the company in 2022. Despite the company’s limited financial resources after its establishment, he successfully developed its own model over a period of more than two years, starting with pre-training through an efficiently designed training pipeline. According to the company, developing the model cost approximately KRW 1.5 billion, while operating costs average around KRW 10 million per month. This is expected to provide a breakthrough solution to the AI cost concerns faced by businesses.
CEO Lee Hwan-ho emphasises that the ‘Ivy’ model offers high efficiency and performance, enabling it to resolve the issue described as a drawback of generative AI: ‘PoC is easy, but completion is not’. This allows businesses to reduce labour costs and improve productivity by having AI perform repetitive and inefficient tasks instead. To minimise AI development costs, CLEVI introduced innovative technology that efficiently manages the development process by applying automated, container-based training node deployment technology. It is also noteworthy that costs were significantly reduced through the efficient design of training pipelines and optimisation of communication networks between nodes.
The ‘Ivy’ model is characterised by the development of a Mixture of models for efficient model training and operation. It was developed by creating small models, then distilling and merging them into a model. In addition, attention and layer expansion were incorporated to develop efficient models through merging or improved routing. Its support for not only English and Korean but also a wide range of languages worldwide through multilingual training and embedding optimisation is another notable feature.
In particular, as OpenAI in the United States and DeepSeek in China compete to advance their AI capabilities, the domestic large-scale generative AI ‘Ivy’ is attracting attention in the industry by recording higher scores than the two existing models in various benchmark tests (△AIME 2024: 81% △MATH-500: 98.6 △SWE-BENCH: 51.1 △MMLU: 91.2).
In addition, as the use of AI Assistants expands and concerns about personal information leaks and security grow, CLEVI has made the Ivy app platform available so that users can register with the minimum amount of personal information. Public institutions and businesses have received positive evaluations for resolving security issues by supporting on-premises installation tailored to their user environments.
A company official said, “The company will not stop at the launch of Ivy, but plans to unveil an agent in the future that directly intervenes, responds and participates in conversations between individual and group users. In addition to enterprise-wide resource planning (ERP) systems, we have also completed R&D on the AI-based proactive workflow agent “ARI” (Adaptive Reasoning Intelligence), which enables more efficient use of AI in various fields, including education and R&D, as well as methods for precise object recognition and real-time control in areas such as video content production and medical image analysis, and are preparing to launch these services. We ask for your continued interest in CLEVI’s activities as it makes an impact in the global AI market through innovative technology, outstanding performance and cost savings.”
Reporter Lee Seung-won [email protected]
Source: Global Economic Daily (Reporter Lee Seung-won) | https://www.getnews.co.kr/news/articleView.html?idxno=813209