All Benchmarks/Multilingual

Horangi 4 (Korean LLM Leaderboard)

Often citedKoreanHigher is better

The overall score of Horangi 4, W&B Korea's Korean-language LLM leaderboard: 20+ Korean tests are combined into General Language Performance (GLP) and Alignment Performance (ALT), and the score is their average. The same model can appear separately for different reasoning-effort settings. Scores run from 0 to 100%. Higher is better.

Other versions:HAE-RAE Bench v1 (Reading)Horangi 4 — ALT (alignment)Horangi 4 — GLP (general language)HRM8KIFEval-KoKMMLUKMMLU-ProKo-AIME 2025Ko-ARC-AGIKo-HLEKo-MoralKo-MTBenchKo-TruthfulQAKoBALT-700 (semantics)KoBALT-700 (syntax)KoBBQ (bias)Korean Hate Speech
Top score84.1%Gemini 3.1 Pro
Models tested49
Last updated2026.08.03
Model release date (newest on the right)
Best score so farHigher on the chart is better
  1. 1
    Google
    Gemini 3.1 Pro
    2026.02.19 · high-effort
    84.1%
  2. 2
    Anthropic
    Claude Fable 5
    2026.06.09 · high-effort
    82.6%
  3. 3
    OpenAI
    GPT-5.5
    2026.04.23 · xhigh-effort
    81.8%
  4. 4
    Google
    Gemini 3.5 Flash
    2026.05.19
    81.2%
  5. 5
    OpenAI
    GPT-5.6 Sol
    2026.07.09 · max-effort
    80.7%
  6. 6
    Alibaba
    Qwen3.7 Max
    2026.05.21
    80.2%
  7. 7
    OpenAI
    GPT-5.4
    2026.03.05 · xhigh-effort
    79.7%
  8. 8
    xAI
    Grok 4.5
    2026.07.08
    78.7%
  9. 9
    Anthropic
    Claude Sonnet 5
    2026.06.30 · high-effort
    78.6%
  10. 10
    OpenAI
    GPT-5.6 Terra
    2026.07.09 · max-effort
    78.3%

The "harness" is the agent program the AI used to carry out the task. The same model can score very differently depending on its harness and reasoning effort.