MGSM
Multilingual Grade School Math - GSM8K translated to 10 languages.
Category
Multilingual
Score type
percent
Scale
100%
higher is better
Models tested
5
What is MGSM?
MGSM (Multilingual Grade School Math) is GSM8K translated into about ten other languages. It shows whether a model's math ability survives when the problem is written in Spanish, German, Bengali, Swahili and so on — a good test of language transfer.
What does it test?
MGSM measures performance in languages other than english. In detail, it checks:
- same math quality across many languages
- understanding of non-English problem statements
- reasoning transfer from English to other languages
- fair treatment of low-resource languages
How to read the score
The score is accuracy: the percentage of tasks answered or solved correctly. 100% means every task was solved perfectly; higher is better.
Frequently asked questions
What is the MGSM benchmark?
MGSM (Multilingual Grade School Math) is GSM8K translated into about ten other languages. It shows whether a model's math ability survives when the problem is written in Spanish, German, Bengali, Swahili and so on — a good test of language transfer.
What does MGSM test?
MGSM mainly measures performance in languages other than English. Concretely, it checks: same math quality across many languages; understanding of non-English problem statements; reasoning transfer from English to other languages.
How is the MGSM score calculated?
The score is accuracy: the percentage of tasks answered or solved correctly. 100% means every task was solved perfectly; higher is better. The declared maximum score of this benchmark is 100.0.
Where do I get MGSM data and scores?
The official source is https://arxiv.org/abs/2210.03057. The official benchmark page/reference is linked on this page. Every score on Open LLM Radar is also available as open data: download the scores JSON (http://open-llm-radar-api:8000/api/v1/benchmarks/mgsm/scores) or pull it directly from the Open LLM Radar API with any HTTP client.
Benchmark data & downloads
Where to get, inspect and reuse the data behind this benchmark.
Official benchmark page / paper
The benchmark maintainer's website, paper or leaderboard
Benchmark metadata (JSON)
This page's benchmark record from the Open LLM Radar API
All model scores (JSON download)
Full score table as JSON; paginate with limit/offset params
Open LLM Radar API docs
Swagger UI to explore and reuse all radar API endpoints
Model scores 5 models tested
| # | Model | Score | Date | Source | Verified |
|---|---|---|---|---|---|
| 1 |
ERNIE-5.5-Flash-Base
Baidu
|
87%
|
— | HF model card baidu/ERNIE-5.5-Flash-Base (EM, 8-shot CoT, Unified Pipeline) | Unverified |
| 2 |
Gemma 3 27B PT
Google DeepMind
|
74.3%
|
27.02.2025 | Report | Verified |
| 3 |
Gemma 3 12B PT
Google DeepMind
|
64.3%
|
27.02.2025 | Report | Verified |
| 4 |
Gemma 3 4B PT
Google DeepMind
|
34.7%
|
27.02.2025 | Report | Verified |
| 5 |
Gemma 3 1B PT
Google DeepMind
|
2.04%
|
27.02.2025 | Report | Verified |