Naver has unveiled a multi-modal inference artificial intelligence (AI) with top-tier language capabilities.
Naver recently announced that it has completed the development of the generative AI 'HyperCLOVA X Think' and published a technical report introducing detailed information such as model design and performance on the 30th.
According to Naver, the results of measuring language capabilities of the large language model (LLM) based on the Korean language performance benchmark 'KoBALT-700' showed that this model achieved higher scores than major domestic inference models built at a similar scale and leading global open-source models.
'HyperCLOVA X Think' ranked first in a total of eight Korean language performance indicators, including 'Haerye' and others, in addition to 'KoBALT-700.'
Naver explained that 'HyperCLOVA X Seed' is a model that combines inference capabilities with multi-modality, allowing it to infer based on visual information as well as language.
In particular, when STEM (science, technology, engineering, and mathematics) problems were input in image format, Naver noted that 'HyperCLOVA X Seed' recognized them and correctly inferred the answers through the inference process.
Naver plans to release the model as open-source. The lightweight model 'HyperCLOVA X Seed,' which Naver open-sourced in April, has surpassed 500,000 cumulative downloads.