AI BriefingOpen Ko-LLM Leaderboard Surpasses 100 Models
Key point
The Open Ko-LLM Leaderboard has surpassed 100 registered models just two weeks after its launch.
Details
The Open Ko-LLM Leaderboard, a Korean LLM evaluation platform built by Upstage and NIA, has surpassed 100 registered models just two weeks after opening. Since its launch on the 27th, participation has continued even through the Chuseok holiday, with daily rankings competition underway.
Participation has expanded widely, from individual researchers to SKT, Maum AI, 42Maru, ETRI, and Korea University. Korean open-source models such as Ko-Alpaca, KULLM, and Polyglot-Ko have also joined, positioning the leaderboard as a benchmark for evaluating Korean-specialized LLM performance.
The leaderboard is an open platform where anyone can register a Korean LLM and compete against other models. It uses high-quality, self-built data reflecting the characteristics and culture of the Korean language, addressing the limitations of English-centric evaluation systems.
Upstage has announced the following upcoming activities:
- Compiling monthly rankings to spotlight top models
- Hosting open seminars inviting developers
- Supporting the growth of the domestic LLM ecosystem
Separately, Upstage's SOLAR ranked No. 1 in the world on the Hugging Face Open LLM Leaderboard, surpassing ChatGPT benchmark results, and has recently drawn global attention by being added to Poe's main model list.
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.