Last Updated: 9 December 2024
Four months after launching EXAONE 3.0, LG AI Research, the dedicated AI think tank of LG Group, has released EXAONE 3.5, its latest generative AI model, as open-source. This step fosters an open AI research ecosystem and is expected to accelerate innovation globally. Positioned among the top-performing AI models in 20 benchmarks, EXAONE 3.5 excels in areas such as real-world usability, long-text processing, coding, and math.
Key advancements in EXAONE 3.5 include:
The model is available in three configurations: an ultra-lightweight model (2.4 billion parameters) for on-device use, a lightweight model (7.8 billion parameters) for general purposes, and a high-performance model (32 billion parameters) for specialized applications.
By making all three models accessible, LG AI Research invites researchers worldwide to validate and innovate.
Coinciding with EXAONE 3.5’s release, LG AI Research has launched ChatEXAONE, an enterprise AI agent designed to revolutionize workplace efficiency. Starting December 9, LG employees can leverage ChatEXAONE for tasks like real-time web searches, document summarization, translation, data analysis, coding, and report writing—all while ensuring data privacy and security.
ChatEXAONE introduces several standout features that set it apart. Its "Deep" function allows for step-by-step analysis to generate detailed responses, while "Dive" lets users customize their search scopes, whether academic, global, or general. Additionally, the tool offers 133 job-specific prompts tailored to 14 job categories, ensuring personalized recommendations.
With scalability supporting up to 32,000 tokens, ChatEXAONE can process extensive inputs of up to 20,000 Korean words or 23,000 English words. Plans are underway to expand this capability to 128,000 tokens in 2025.
This integration marks the beginning of an "era of everyday AI," empowering employees to enhance productivity securely and efficiently.
LG AI Research celebrated its fourth anniversary on December 7, 2024, by reaffirming its commitment to innovation and collaboration. The organization is actively developing a Large Action Model (LAM) for autonomous planning and execution, targeting a 2025 launch for advanced AI agents.
Collaboration remains central to LG’s strategy. Partnerships with leading institutions such as AWS, NVIDIA, Intel, and academic centers like the University of Toronto aim to create real-world applications and outcomes for EXAONE technologies.
“The recent advancement of generative AI models has accelerated, and it is important to upgrade them quickly. We will speed up the pace of innovation and develop them into frontier models that represent Korea, with the goal of artificial super intelligence that can be applied to real-world industries.”
Bae Kyunghoon, president of LG AI Research.
As part of the LG Group’s long-term vision, LG AI Research prioritizes AI transformation and sustainable growth. Chairman Koo Kwang-mo’s focus on future technologies highlights LG’s proactive efforts to recruit top talent and pioneer advancements in AI, bio, and cleantech. Through initiatives like EXAONE 3.5 and beyond, LG continues to shape a future where AI seamlessly integrates into everyday life.
Source: PR Newswire
