AI BriefingLG AI Research: EXAONE 3.0 Released
Key point
LG AI Research unveiled EXAONE 3.0 7.8B, an open-source model with excellent English and Korean performance.
Details
After 3 years of research, LG AI Research has open-sourced the EXAONE 3.0 7.8B Instruction Tuned model. This model is based on the latest trend, the Decoder-only Transformer Architecture, and uses 7.8B parameters with 8T tokens of training data.
Key performance characteristics are as follows:
- English Performance: It recorded global top-tier results in real-world use cases and benchmarks, achieving particularly outstanding scores in math, coding, and reasoning.
- Korean Performance: It ranked #1 overall in both Korean real-world use cases and general performance benchmarks, demonstrating excellent bilingual capability.
- Cost Efficiency: Through 3 years of R&D, costs were significantly reduced to 6% compared to the initial model, increasing its practical applicability in industry settings.
This model aims to help researchers advance the AI ecosystem and contribute to building Expert AI that can be utilized in specialized industrial fields.
This summary was generated automatically by AI. Check the original for the author's claims and context. Copyright belongs to the original author.
Our guide explains how the AI works. Report summary errors, attribution issues, or removal requests via Contact.
Related
Qwen3.5-based Small Decision Model 'Kev' Released with TypeSafe API Compatibility and Performance Metrics