Benchmark Almanac

KorNAT

Claude Opus 5 65.4% leads since July 2026
Not in model cards

Models answer questions about Korean social values and common knowledge. The authors compare answers with Korean reference judgments.

The Open Ko-LLM Leaderboard reports knowledge and social value scores separately.

Made by
Jiyoung Lee and colleagues
Released
February 2024
Kind
Knowledge in many languages
Level
Everyday
Scored
Checked automatically
Answers
Multiple choice
Tasks
Public, plus a private test set
Measures
Capability
Language
Korean
In and out
Text → Text
Interaction
Single turn
Refresh
One fixed set
Run by
A host
Strongest model tested
Claude Opus 5 #11 on the capability index
Last model added
August 2026
Scores refreshed
November 20th, 2026
20
Entries on its own leaderboard, 17 models we track

Claude Opus 5 leads with 65.4% among models released by August 2026 Claude Opus 5 leads with 65.4% among models released by August 2026

Best score per entry from the benchmark's own leaderboard, read on November 20th, 2026. The leaderboard reports no overall score, so this is the plain mean of Social Values and Common Knowledge. Where a model ran with several settings or harnesses, we keep its best. Names in gray aren't in our model list: fine-tunes, agent systems, products and older models. They don't count toward leaders. This view shows only models we track, each at its best.

The newest model with a score here came out in August 2026, and nobody has tested a newer one. Among the 20 highest-ranked models on the capability index, these have no score yet:

  1. 1 Claude Opus 5 ✓ 65.4%
  2. 2 GPT-5.6 Sol ✓ 64.8%
  3. 3 Solar Pro 4 ✓ 63.4%
  4. 4 Qwen 3.8 Max ✓ 63.2%
  5. 5 Kimi K3 ✓ 62.5%
  6. 6 Grok 4.6 ✓ 62%
  7. 7 GPT-5.6 Luna ✓ 61.9%
  8. 8 naver-hyperclovax/HyperCLOVAX-SEED-Think-32B ✓ 60.5%
  9. 9 Gemini 3.1 Pro ✓ 60%
  10. 10 EXAONE 4.0 32B ✓ 59%
  11. 11 Gemma 3 27B ✓ 58.4%
  12. 12 GLM-5.2 ✓ 58.2%
  13. 13 DeepSeek-V4-Pro ✓ 57.5%
  14. 14 Qwen3-32B ✓ 57.4%
  15. 15 Gemini 3.6 Flash ✓ 57.4%
  16. 16 EXAONE 4.5 33B ✓ 57%
  17. 17 Llama 3.3 70B Instruct ✓ 56.4%
  18. 18 Claude Haiku 4.5 ✓ 56.1%
  19. 19 skt/A.X-4.0-Light ✓ 50.8%
  20. 20 kakaocorp/kanana-1.5-8b-instruct-2505 ✓ 48.6%

Source: Official KorNAT leaderboard

#EntryScoreSettingSubmitted byReleasedRun onSource
1 Claude Opus 5 ✓ 65.4% organizers Jul 2026 Aug 2026 Official leaderboard
2 GPT-5.6 Sol ✓ 64.8% organizers Jul 2026 Aug 2026 Official leaderboard
3 Solar Pro 4 ✓ 63.4% organizers Aug 2026 Aug 2026 Official leaderboard
4 Qwen 3.8 Max ✓ 63.2% organizers Aug 2026 Aug 2026 Official leaderboard
5 Kimi K3 ✓ 62.5% organizers Jul 2026 Aug 2026 Official leaderboard
6 Grok 4.6 ✓ 62% organizers Aug 2026 Aug 2026 Official leaderboard
7 GPT-5.6 Luna ✓ 61.9% organizers Jul 2026 Aug 2026 Official leaderboard
8 naver-hyperclovax/HyperCLOVAX-SEED-Think-32B ✓ 60.5% organizers Aug 2026 Official leaderboard
9 Gemini 3.1 Pro ✓ 60% organizers Feb 2026 Aug 2026 Official leaderboard
10 EXAONE 4.0 32B ✓ 59% organizers Jul 2025 Aug 2026 Official leaderboard
11 Gemma 3 27B ✓ 58.4% organizers Mar 2025 Aug 2026 Official leaderboard
12 GLM-5.2 ✓ 58.2% organizers Jun 2026 Aug 2026 Official leaderboard
13 DeepSeek-V4-Pro ✓ 57.5% organizers Apr 2026 Aug 2026 Official leaderboard
14 Qwen3-32B ✓ 57.4% organizers Apr 2025 Aug 2026 Official leaderboard
15 Gemini 3.6 Flash ✓ 57.4% organizers Jul 2026 Aug 2026 Official leaderboard
16 EXAONE 4.5 33B ✓ 57% organizers Apr 2026 Aug 2026 Official leaderboard
17 Llama 3.3 70B Instruct ✓ 56.4% organizers Dec 2024 Aug 2026 Official leaderboard
18 Claude Haiku 4.5 ✓ 56.1% organizers Oct 2025 Aug 2026 Official leaderboard
19 skt/A.X-4.0-Light ✓ 50.8% organizers Aug 2026 Official leaderboard
20 kakaocorp/kanana-1.5-8b-instruct-2505 ✓ 48.6% organizers Aug 2026 Official leaderboard
#ModelScoreSettingSubmitted byReleasedRun onSource
1 Claude Opus 5 ✓ 65.4% organizers Jul 2026 Aug 2026 Official leaderboard
2 GPT-5.6 Sol ✓ 64.8% organizers Jul 2026 Aug 2026 Official leaderboard
3 Solar Pro 4 ✓ 63.4% organizers Aug 2026 Aug 2026 Official leaderboard
4 Qwen 3.8 Max ✓ 63.2% organizers Aug 2026 Aug 2026 Official leaderboard
5 Kimi K3 ✓ 62.5% organizers Jul 2026 Aug 2026 Official leaderboard
6 Grok 4.6 ✓ 62% organizers Aug 2026 Aug 2026 Official leaderboard
7 GPT-5.6 Luna ✓ 61.9% organizers Jul 2026 Aug 2026 Official leaderboard
8 Gemini 3.1 Pro ✓ 60% organizers Feb 2026 Aug 2026 Official leaderboard
9 EXAONE 4.0 32B ✓ 59% organizers Jul 2025 Aug 2026 Official leaderboard
10 Gemma 3 27B ✓ 58.4% organizers Mar 2025 Aug 2026 Official leaderboard
11 GLM-5.2 ✓ 58.2% organizers Jun 2026 Aug 2026 Official leaderboard
12 DeepSeek-V4-Pro ✓ 57.5% organizers Apr 2026 Aug 2026 Official leaderboard
13 Qwen3-32B ✓ 57.4% organizers Apr 2025 Aug 2026 Official leaderboard
14 Gemini 3.6 Flash ✓ 57.4% organizers Jul 2026 Aug 2026 Official leaderboard
15 EXAONE 4.5 33B ✓ 57% organizers Apr 2026 Aug 2026 Official leaderboard
16 Llama 3.3 70B Instruct ✓ 56.4% organizers Dec 2024 Aug 2026 Official leaderboard
17 Claude Haiku 4.5 ✓ 56.1% organizers Oct 2025 Aug 2026 Official leaderboard

Rankings by part

We read 9 parts from the benchmark’s own leaderboard, each ranking models on one part of the test, such as Social Values, Common Knowledge and Common Sense. Gemini 3.1 Pro leads 8 of them, and EXAONE 4.0 32B leads the other one. Arrows show how far a model moved from its overall rank among the models we track.

Social Values

17 models we track have a score here.

  1. 1 EXAONE 4.0 32B ↑8 33.3%
  2. 2 Claude Opus 5 ↓1 33%
  3. 3 GPT-5.6 Sol ↓1 32%
  4. 4 Qwen 3.8 Max 30.6%
  5. 5 Solar Pro 4 ↓2 30.4%
  6. 6 Gemma 3 27B ↑4 30.3%
  7. 7 Kimi K3 ↓2 29.9%
  8. 8 Grok 4.6 ↓2 29.2%
  9. 9 GPT-5.6 Luna ↓2 29.1%
  10. 10 Qwen3-32B ↑3 29%

6 models have held number one on the official leaderboard since its release in February 2024

Lab claims count from the day the card came out, independent runs from the model’s release.

Meta Google DeepMind LG AI Research OpenAI Anthropic
Own leaderboard Not tested since Aug 2026

Source: Official KorNAT leaderboard

Scores over time

All the scores we have, one dot each, so a model tested at several settings appears more than once.

Plot with other benchmarks
On the benchmark’s own leaderboard Best so far
0% 20% 40% 60% 80% 100% 2025 2026 Near the ceiling (85%) Claude Opus 5 Gemini 3.1 Pro EXAONE 4.0 32B Llama 3.3 70B Instruct 0% 20% 40% 60% 80% 100% 2025 2026 Near the ceiling (85%) Claude Opus 5 EXAONE 4.0 32B

Source: Official KorNAT leaderboard

No lab model card reports it yet

We read the headline results of 540 launch posts from 57 labs, and none of them list this benchmark.

More in Multilingual

All 545 in Multilingual
  • MMMLU In use MMLU professionally translated into 14 languages. Multiple-choice accuracy. Claude Mythos Preview 92.7%
  • WMT24++ In use English-to-55-language translation of WMT24 test data. Scored with automatic metrics such as MetricX or xCOMET. K-EXAONE 90.5%
  • MMLU-ProX In use MMLU-Pro translated into up to 29 languages with expert review. Multiple-choice accuracy per language. Qwen3.7-Max 87%
  • CLIcK In use Models answer 1,995 multiple-choice questions on Korean language and culture. Questions come from official Korean exams and textbooks in eleven categories. Solar Open 2 90.7%
  • Global MMLU Lite 3.0 In use Models answer 400 human-translated MMLU questions per language in 23 languages. Version 3.0 adds Czech, Hungarian, Oriya, Slovak and Tajik and updates Italian. Accuracy. Claude Opus 5 95.7%
  • MMMB In use Models answer multiple-choice questions about images in six languages across 15 categories. The set has about 12,000 questions. LFM2.5-VL-3B 83%

Data audit

All 10 checks pass. Every benchmark

Catalog record 2 pass
  • Pass Task access Public tasks, plus a private test set.
  • Pass Confidence Checked against a primary source.
Scores 4 pass 1 note
  • Pass Independent runs Scores from the official leaderboard (20 entries).
  • Pass Models scored 17 models we track, plus 3 other entries on its leaderboard.
  • Pass Newest score The newest is from August 2026, for Grok 4.6.
  • Note Frontier models None of the top 10 has a score. The strongest tested is Claude Opus 5 (#11).
  • Pass Headroom The best is 65.4%, under 85%, so it still separates models.
Leaderboard import 3 pass
  • Pass Importer We read the board on huggingface.co: 17 rows for models we track and 3 other entries at the last read.
  • Pass Last read November 20th, 2026.
  • Pass Read worked The last read worked.
Consistency 1 pass 1 note
  • Note Lab claims No model has both a lab number and an independent run.
  • Pass Known issues No published audit found problems.

How we check