Naver HyperCLOVA X Open Source Surpasses 300,000 Downloads in a Month

Naver Cloud's three open-source AI models, 'HyperCLOVA X Seed,' surpassed 300,000 downloads in just over a month since their launch, astonishing the global AI community. This record places them in the top approximately 0.03% of models registered on Hugging Face, a global AI model sharing community, clearly demonstrating HyperCLOVA X's expanding global influence. In particular, the visual-language model 'HyperCLOVA X Seed 3B,' which processes text, image, and video information together, garnered the most significant response, recording approximately 267,000 downloads. This model scored higher than similarly sized U.S. and Chinese big tech models in Korean and English visual information understanding benchmarks, and demonstrated comparable accuracy rates to much larger overseas models, proving its excellent performance competitiveness. The text models 'HyperCLOVA X Seed 1.5B' and '0.5B' also recorded approximately 35,000 downloads, and active improvement work and experiments are ongoing within the developer community, with as many as 29 derived models being developed based on these two. Developers provided positive feedback, stating, "Despite their small size, they stand out with balanced language performance and excellent stability," and "You can use a smart Korean model on-device with only about 1GB of RAM." Furthermore, practical application reviews were shared, such as "It responds excellently even with just SFT+LoRA training and is fast even in a CPU environment, making it easy to create an in-house chatbot," hinting at its high potential in real business environments. Seong Nak-ho, Head of Hyperscale AI Technology at Naver Cloud, stated, "It is very meaningful that generative AI models developed by Team Naver from the ground up are gaining attention in the global open-source community," adding, "We hope HyperCLOVA X can contribute to adding diversity to the AI ecosystem and fostering even more innovative technologies and services based on it." Naver Cloud plans to unveil the HyperCLOVA X inference model currently under development through its conversational AI service 'CLOVA X' in June. Furthermore, it plans to expand multi-modality beyond text, image, and video to include voice, sequentially launching services utilizing HyperCLOVA X voice models, thereby strengthening its leadership in the AI field.