South Korean giant Naver has announced a trimmed-down version of its own HyperCLOVA X artificial intelligence model. This time, the version is actually 40% smaller in size, and yet it performs much better than that of its predecessors-another breakthrough in AI technology.
Enhanced Performance
Despite this reduced size, HyperCLOVA X outperforms its predecessor on 19 benchmark tests-some of which are Korean, English, coding, and mathematics. It posted a 79.6% accuracy score in the Massive Multitask Language Understanding (MMLU) benchmark. Which was compared with top-tier AI models such as Anthropic’s Claude 3 Sonnet and Alibaba’s Qwen-32B.
Multimodal Capabilities
HyperCLOVA X has now become better purposed for a multimodal usage, which allows it to be trained and made to process and understand images and even videos. This process facilitated more powerful question-answering systems concerning pictures and the more sophisticated understanding of charts and graphs. Thereby, making it more general as far as AI applications are concerned.
Operational Efficiency
By improving the structure and reducing parameter size, Naver improved HyperCLOVA X’s operational efficiency by more than 50%. The efficiency leap results in reduced operational costs. But also at the same time in speeding up the process of integrating AI in line with various services.
Integration and Future Plans
Naver aims to incorporate HyperCLOVA X into its conversational AI chatbot CLOVA X before March. Enterprise customers will have access to the model through the Naver Cloud Platform. Naver intends to improve AI planning and reasoning capabilities and plans to release a conversational AI version of HyperCLOVA X by the second half of 2025.
