LiveK12Bench - Mathematics Exam Score: leaderboard
Overall exam score (0-100) on the mathematics questions of LiveK12Bench (2026), recent Chinese high-school examination papers graded as a human marker would: a judged score for the reasoning process combined with the outcome score of each question. Higher is better.
Metric: Overall exam score (0-100), human-style grading that weights a judged process score and the outcome score of each question, Mathematics questions, LiveK12Bench 2026-03 split of recent Chinese high-school exam papers, unified answer prompt, answers checked by a panel of three LLM judges that never grade their own model; higher is better. Source: arxiv.org. Saturation forecast: Estimated already saturated. 12 models tracked. Updated 7 Oct 2026.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemini 3 Pro | 90.3 | #146 |
| 2 | Claude Opus 4.6 | 90 | #114 |
| 3 | Gemini 3 Flash | 89.8 | #180 |
| 4 | Kimi K2.5 | 87.8 | #272 |
| 5 | Claude Sonnet 4.6 | 87.4 | #179 |
| 6 | GPT-5 | 85.5 | #185 |
| 7 | Qwen 3 VL 235B A22B | 81.3 | #486 |
| 8 | Qwen 3 VL 32B | 81.1 | #624 |
| 9 | GPT-5 Mini | 79.8 | #373 |
| 10 | GLM-5 | 79.5 | #276 |
| 11 | Qwen 3 VL 8B | 77.2 | #927 |
| 12 | GPT-4o | 24.2 | #754 |
No result here: #2 Claude Fable 5.2, #3 Gemini 4 Neon, #4 GPT-6.2 Astra.
Interactive version: theaggregate.ai/benchmark?slug=livek12bench-mathematics-exam-score · How It Works · Data refreshed daily, snapshot 2026-11-22.