All Benchmarks/Math

Ko-AIME 2025

KoreanHigher is better

The 30 problems of the 2025 AIME translated into Korean and checked by people. Answers are integers, and with only 30 problems, each one is worth about 3.3 percentage points. 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-ARC-AGIKo-HLEKo-MoralKo-MTBenchKo-TruthfulQAKoBALT-700 (semantics)KoBALT-700 (syntax)KoBBQ (bias)Korean Hate Speech
Top score100.0%GPT-5.5
Models tested48
Last updated2026.08.03
Model release date (newest on the right)
Best score so farHigher on the chart is better
  1. 1
    OpenAI
    GPT-5.5
    2026.04.23 · xhigh-effort
    100.0%
  2. 1
    Anthropic
    Claude Opus 4.8
    2026.05.27 · high-effort
    100.0%
  3. 1
    OpenAI
    GPT-5.4 Nano
    2026.03.17 · xhigh-effort
    100.0%
  4. 1
    Anthropic
    Claude Fable 5
    2026.06.09 · high-effort
    100.0%
  5. 1
    xAI
    Grok 4.5
    2026.07.08
    100.0%
  6. 1
    OpenAI
    GPT-5.4
    2026.03.05 · xhigh-effort
    100.0%
  7. 7
    Google
    Gemini 3 Flash
    2025.12.17 · high-effort
    96.7%
  8. 7
    OpenAI
    GPT-5.6 Luna
    2026.07.09 · max-effort
    96.7%
  9. 7
    OpenAI
    GPT-5 Mini
    2025.08.07
    96.7%
  10. 7
    NVIDIA
    Nemotron 3 Super
    2026.03.11
    96.7%

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.