Naver Cloud Builds Finance-Specialized sLLM for Mirae Asset Securities

Provided as a hyperclovaX-based sLLM construction type

Naver Cloud announced on the 19th that it has built a small language model (sLLM) specialized for the financial environment based on the lightweight model 'HyperClovaX Dash (hereinafter HCX-DASH)' in Mirae Asset Securities' on-premises (IT infrastructure built independently by the company) environment.

Naver Cloud's lightweight model 'HyperClovaX Dash'

Naver Cloud has started a build-type large language model (LLM) business to enable public and financial sector companies, which are reluctant to adopt cloud-based generative artificial intelligence (AI) due to security issues, to utilize generative AI in on-premises environments. Mirae Asset Securities became the first case.

The build-type LLM is a model implemented within the IT infrastructure independently owned by the company, not on the public cloud. Naver Cloud provided the HCX-DASH model along with class code and technical support to train the data owned by Mirae Asset Securities. Based on this, Mirae Asset Securities trained internal data to build an sLLM optimized for financial tasks.

Im Tae-geon, Executive Director of Naver Cloud, said, "By offering build-type models that can be installed and used on corporate infrastructure, we have strengthened our lineup so that clients can choose and use the form they want according to their purpose and environment for using generative AI. We expect this to be an opportunity for public and financial companies, which had difficulties using cloud-based LLMs due to security reasons, to utilize generative AI."