# KMMLU leaderboard

korean benchmark

> Every figure below is reproduced as its upstream source published it: nothing is modelled, estimated, interpolated or converted. `Unavailable` means no source published the value — it is never a zero. Each value carries its evidence state and the date it was observed.

Page: https://modelscale.dev/leaderboards/benchmarks/korean-kmmlu

Korean Massive Multitask Language Understanding. Every model the catalog carries a published KMMLU value for, ranked by that value.

## Benchmark record

| Field | Value |
| --- | --- |
| Category | korean |
| Measure | Multiple choice questions |
| Tasks | 35,030 questions |
| Difficulty | Elementary to professional level in Korean |
| Published by | [KMMLU: Measuring Massive Multitask Language Understanding in Korean](https://arxiv.org/abs/2402.11548) |

Evaluates Korean expert-level knowledge across 45 subjects. 20% of questions require Korean cultural context.

## KMMLU ranking

2 models with a published KMMLU value, ordered by that value, highest first. Models the source has not scored on this benchmark are not listed — they are unmeasured, not last.

| Rank | Model | Provider | Multiple choice questions | Evidence |
| --- | --- | --- | --- | --- |
| 1 | [Kanana-2 3B Instruct](https://modelscale.dev/models/kanana-2-3b-instruct) | Kakao | 43.32 | Observed 2026-11-21 · source benchlm:models |
| 2 | [Kanana-2 1.3B Instruct](https://modelscale.dev/models/kanana-2-1-3b-instruct) | Kakao | 42.79 | Observed 2026-11-21 · source benchlm:models |

Rows are ordered by the value the source published, highest first. The source does not state whether a higher value is the better result, so this document does not either: for a benchmark that measures a rate of failure, read the table from the bottom.
