LG unveils upgraded version of Exaone AI model with advanced features

LG AI Research, the artificial intelligence division of South Korea's LG Group, has unveiled an enhanced version of its multimodal AI model, Exaone 3.5. This upgraded model promises improved performance and greater cost efficiency for users.

To foster innovation in the AI ecosystem, LG announced that all three variants of Exaone 3.5 will be released as open source, making them available for a variety of research applications. This move follows the open-source release of the lightweight version of its predecessor, Exaone 3.0, just four months ago, according to Yonhap news agency.

The Exaone 3.5 incorporates advanced technologies designed to enhance accuracy and reliability by generating answers based on real-time web searches or uploaded documents. This reduces the likelihood of plausible yet incorrect responses.

In recent evaluations, the upgraded model outperformed other global open-source AI models from the U.S., China, and beyond, earning top scores in areas such as long-text processing, coding, mathematics, and practical usability.

“As AI becomes a critical strategic asset for nations, developing models with homegrown technology is vital for boosting national AI competitiveness,” stated an official from LG AI Research.

Exaone, short for "Expert AI for Everyone," is a multimodal language model capable of understanding and generating both text and images simultaneously, enabling image-to-text and text-to-image tasks. Its deep learning algorithm processes vast amounts of data to provide solutions and address complex questions across various fields.

The latest release builds on LG’s 2022 unveiling of the second-generation Exaone, cementing its commitment to becoming a key player in the global AI race. LG AI Research, established in 2020, is backed by LG Group’s ambitious plan to invest 3.6 trillion won ($2.8 billion) in AI research and development over five years.