LatamBoard - Spanish MGSM: leaderboard
The "Spanish MGSM" subtask of LatamBoard. Leaderboard for model performance on Latin American and Iberian language tasks, including Spanish and Portuguese understanding, TELEIA Spanish exams, FLORES/OPUS translation, and structured or image extraction.
Metric: Score (%). Source: huggingface.co. Saturation forecast: Not forecast. 35 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | gemma-3-12B-pt | 30.8 | #686 |
| 2 | Phi-3-medium-4k-instruct | 27.6 | #952 |
| 3 | Yi-1.5-9B Chat | 18.8 | #1031 |
| 4 | Phi-4 Multimodal Instruct | 15.2 | #906 |
| 5 | Phi-4 Mini Instruct | 14.4 | #1245 |
| 6 | Qwen 3 VL 8B Instruct | 14 | #420 |
| 7 | Granite 4.0 H 1B | 12.8 | #1191 |
| 8 | Gemma 3n E4B (IT) | 12.4 | #749 |
| 9 | Mistral-NeMo-Minitron-8B-Instruct | 12.4 | #907 |
| 10 | aya-vision-8B | 12 | #1107 |
| 11 | Hermes-4-14B | 11.6 | #751 |
| 12 | SOLAR-10.7B-Instruct-v1.0 | 10.4 | #1075 |
| 13 | Yi-1.5-6B-Chat | 9.6 | #1259 |
| 14 | DeepSeek-R1-Distill-Qwen-7B | 8.4 | #1415 |
| 15 | aya-expanse-8B | 8 | #931 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=latamboard-spanish-mgsm · How It Works · Data refreshed daily, snapshot 2026-10-20.