All Benchmarks/Multilingual

KoBALT-700 (semantics)

KoreanHigher is better

The semantics part of KoBALT-700, a set of Korean linguistics questions written by experts: multiple-choice items on meaning, such as whether a predicate and its arguments fit together semantically. Horangi runs a fixed 100-question subset rather than the full dataset, so its numbers are not directly comparable with full-dataset scores. 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)Horangi 4 (Korean LLM Leaderboard)HRM8KIFEval-KoKMMLUKMMLU-ProKo-AIME 2025Ko-ARC-AGIKo-HLEKo-MoralKo-MTBenchKo-TruthfulQAKoBALT-700 (syntax)KoBBQ (bias)Korean Hate Speech
Top score91.0%Gemini 3.1 Pro
Models tested48
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
    91.0%
  2. 2
    Anthropic
    Claude Fable 5
    2026.06.09 · high-effort
    87.0%
  3. 3
    OpenAI
    GPT-5.6 Terra
    2026.07.09 · max-effort
    84.0%
  4. 3
    OpenAI
    GPT-5.6 Sol
    2026.07.09 · max-effort
    84.0%
  5. 5
    OpenAI
    GPT-5.5
    2026.04.23 · xhigh-effort
    83.0%
  6. 6
    Google
    Gemini 3.5 Flash
    2026.05.19
    82.8%
  7. 7
    Google
    Gemini 3 Flash
    2025.12.17 · high-effort
    82.0%
  8. 7
    Z.ai
    GLM-5
    2026.02.11
    82.0%
  9. 9
    Alibaba
    Qwen3.7 Max
    2026.05.21
    81.3%
  10. 10
    xAI
    Grok 4.5
    2026.07.08
    80.6%

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.