Kaggle Open Benchmarks - MGSM: leaderboard
Average accuracy on MGSM, grade-school math word problems translated into 11 languages (Bengali, Chinese, English, French, German, Japanese, Russian, Spanish, Swahili, Telugu, Thai); published only for models with results in all 11 languages. Kaggle's standardized re-run on its own harness under the open-benchmarks organization, not the original authors' leaderboard; Kaggle's fraction is shown as a percentage (0-100, higher is better).
Metric: Accuracy (%). Source: www.kaggle.com. Saturation forecast: Estimated already saturated. 44 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Claude Opus 4.8 | 95.45 | #39 |
| 2 | Gemini 3.5 Flash | 94.46 | #66 |
| 3 | Claude Opus 4 | 94.36 | #170 |
| 4 | Claude Opus 4.1 | 94 | #146 |
| 5 | Gemini 3 Pro (Preview) | 93.71 | #71 |
| 6 | O4 Mini | 93.35 | #185 |
| 7 | Claude 3.5 Sonnet (20241022) | 92.95 | #302 |
| 8 | GPT-5 | 92.65 | #105 |
| 9 | Gemma 4 26B A4B | 92.64 | #360 |
| 10 | Claude 3.7 Sonnet | 92.58 | #259 |
| 11 | Gemini 2.5 Pro (Preview 06-05) | 92.58 | #118 |
| 12 | Claude Sonnet 4 | 92.44 | #207 |
| 13 | DeepSeek V3 | 92 | #329 |
| 14 | DeepSeek R1 0528 | 91.99 | #232 |
| 15 | O3 | 91.53 | #132 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=kaggle-open-benchmarks-mgsm · How It Works · Data refreshed daily, snapshot 2026-11-05.