Math-PT (pt-BR MC) - Level 3: leaderboard

Math-PT (2026) tests mathematical reasoning on problems written natively in Brazilian Portuguese; this board covers the level-3 multiple-choice questions from OBMEP. Accuracy in percent on the boxed answer letter, higher is better.

Metric: Accuracy (%) on the 140 level-3 multiple-choice questions in Brazilian Portuguese (OBMEP); native-language zero-shot prompt asking for a boxed final answer, figures given as LaTeX source; the letter in the final boxed answer must match the key; higher is better. Source: arxiv.org. Saturation forecast: Estimated already saturated. 11 models tracked. Updated 7 Oct 2026.

Top models

#ModelScoreOverall rank
1GPT-592.86#178
2Qwen 3 235B A22B92.86#628
3DeepSeek V3.191.43#511
4Qwen 3 14B90#1269
5Qwen 3 32B89.29#972
6Qwen 3 30B A3B89.29#1156
7Gemini 2.5 Flash88.57#463
8Qwen 3 8B87.14#1873
9Claude Haiku 4.584.29#583
10Gemma 3 27B (IT)69.29#1324
11Llama 3.3 70B Instruct35#1385

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

Interactive version: theaggregate.ai/benchmark?slug=math-pt-pt-br-mc-level-3 · How It Works · Data refreshed daily, snapshot 2026-11-21.