K12-Bench - Mathematics: leaderboard

The mathematics subset of K12-Bench (2026), all five curriculum tasks (knowledge grounding, prerequisite reasoning, neighbour recommendation, experiment evidence, cross-chapter indexing) built from the K12-KGraph knowledge graph of Chinese K-12 mathematics, physics, chemistry and biology textbooks, as four-option multi-select questions answered zero-shot. Exact match in percent (every option must be chosen correctly); higher is better.

Metric: Mathematics subset, exact match (%, all-or-nothing over the selected option set) on four-option multi-select questions generated from the K12-KGraph curriculum knowledge graph of Chinese K-12 mathematics, physics, chemistry and biology textbooks; zero-shot, answer-only, temperature 0, all five tasks; higher is better. Source: arxiv.org. Saturation forecast: Around February 2027. 10 models tracked. Updated 7 Oct 2026.

Top models

#ModelScoreOverall rank
1Gemini 3 Flash59.8#175
2Gemini 2.5 Flash50.8#463
3Gemma 4 31B (IT)47.8#350
4GPT-5.244.8#185
5Qwen 3 32B44.2#972
6Ministral-3-14B-Instruct38.9#1139
7GLM-4.7 Flash34#1349
8GPT-5 Mini32.4#364
9GPT-4o31.9#751
10Llama 3 8B Instruct5.2#4852

No result here: #1 Claude Fable 5.2, #3 Gemini 4 Neon, #4 GPT-6.2 Astra.

Interactive version: theaggregate.ai/benchmark?slug=k12-bench-mathematics · How It Works · Data refreshed daily, snapshot 2026-11-21.