All Benchmarks/Math

HRM8K

KoreanHigher is better

Math word problems in Korean, combining problems from Korean exams, such as the Korean Mathematical Olympiad and the national college entrance exam (CSAT), with translated problems from well-known English math sets. Final answers are checked by a rule-based grader. 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)IFEval-KoKMMLUKMMLU-ProKo-AIME 2025Ko-ARC-AGIKo-HLEKo-MoralKo-MTBenchKo-TruthfulQAKoBALT-700 (semantics)KoBALT-700 (syntax)KoBBQ (bias)Korean Hate Speech
Top score97.0%Gemini 3 Flash
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 Flash
    2025.12.17 · high-effort
    97.0%
  2. 1
    Anthropic
    Claude Fable 5
    2026.06.09 · high-effort
    97.0%
  3. 1
    Alibaba
    Qwen3.7 Max
    2026.05.21
    97.0%
  4. 4
    OpenAI
    GPT-5.4 Nano
    2026.03.17 · xhigh-effort
    96.0%
  5. 4
    Google
    Gemini 3.5 Flash
    2026.05.19
    96.0%
  6. 4
    OpenAI
    GPT-5.4 Mini
    2026.03.17 · xhigh-effort
    96.0%
  7. 7
    DeepSeek
    DeepSeek V3.2
    2025.12.01 · high-effort
    95.8%
  8. 8
    Anthropic
    Claude Opus 4.8
    2026.05.27 · high-effort
    95.0%
  9. 8
    Google
    Gemini 3.1 Pro
    2026.02.19 · high-effort
    95.0%
  10. 8
    LG AI Research
    K-EXAONE
    2026.01.12 · thinking
    95.0%

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.