MGSM
MGSM (Multilingual Grade School Math) is a benchmark of grade-school math problems. Contains 250 grade-school math problems manually translated from the GSM8K dataset into ten typologically diverse languages: Spanish, French, German, Russian, Chinese, Japanese, Thai, Swahili, Bengali, and Telugu. Evaluates multilingual mathematical reasoning capabilities.
Progress Over Time
Interactive timeline showing model performance evolution on MGSM
State-of-the-art frontier
Open
Proprietary
MGSM Leaderboard
31 models • 0 verified
| Context | Cost | License | ||||
|---|---|---|---|---|---|---|
1 |  Meta | 0.923 | 400B | 1.0M | $0.17 $0.85 | |
2 |  OpenAI | 0.920 | — | 200K | $1.10 $4.40 | |
3 |  Anthropic | 0.916 | — | 200K | $3.00 $15.00 | |
3 | Anthropic | 0.916 | — | 200K | $3.00 $15.00 | |
5 | 0.911 | 70B | 128K | $0.20 $0.20 | ||
6 | OpenAI | 0.908 | — | 128K | $15.00 $60.00 | |
7 | Anthropic | 0.907 | — | 200K | $15.00 $75.00 | |
8 | Meta | 0.906 | 109B | 10.0M | $0.08 $0.30 | |
9 | OpenAI | 0.905 | — | 128K | $2.50 $10.00 | |
10 | OpenAI | 0.893 | — | 200K | $15.00 $60.00 | |
11 | OpenAI | 0.885 | — | 128K | $10.00 $30.00 | |
12 |  Google | 0.875 | — | 2.1M | $2.50 $10.00 | |
13 | OpenAI | 0.870 | — | 128K | $0.15 $0.60 | |
14 | 0.869 | 90B | 128K | $0.35 $0.40 | ||
15 | Anthropic | 0.856 | — | 200K | $0.80 $4.00 | |
16 |  Alibaba Cloud / Qwen Team | 0.835 | 235B | 128K | $0.10 $0.10 | |
17 | Anthropic | 0.835 | — | 200K | $3.00 $15.00 | |
18 | Google | 0.826 | — | 1.0M | $0.15 $0.60 | |
19 |  Microsoft | 0.806 | 15B | 16K | $0.07 $0.14 | |
20 | Anthropic | 0.751 | — | 200K | $0.25 $1.25 | |
21 | OpenAI | 0.745 | — | 33K | $30.00 $60.00 | |
22 | 0.689 | 11B | 128K | $0.05 $0.05 | ||
23 | Google | 0.670 | 8B | 32K | $20.00 $40.00 | |
24 | Microsoft | 0.639 | 4B | — | — | |
25 | 0.607 | 2B | — | — | ||
26 | Microsoft | 0.587 | 60B | — | — | |
27 | 0.582 | 3B | 128K | $0.01 $0.02 | ||
28 | OpenAI | 0.563 | — | 16K | $0.50 $1.50 | |
29 | 0.531 | 2B | — | — | ||
29 | Google | 0.531 | 8B | — | — | |
31 | Microsoft | 0.479 | 4B | 128K | $0.10 $0.10 |
Notice missing or incorrect data?Start an Issue discussion→
FAQ
Common questions about MGSM
MGSM (Multilingual Grade School Math) is a benchmark of grade-school math problems. Contains 250 grade-school math problems manually translated from the GSM8K dataset into ten typologically diverse languages: Spanish, French, German, Russian, Chinese, Japanese, Thai, Swahili, Bengali, and Telugu. Evaluates multilingual mathematical reasoning capabilities.
The MGSM paper is available at https://arxiv.org/abs/2210.03057. This paper provides detailed information about the benchmark methodology, dataset creation, and evaluation criteria.
The MGSM leaderboard ranks 31 AI models based on their performance on this benchmark. Currently, Llama 4 Maverick by Meta leads with a score of 0.923. The average score across all models is 0.779.
The highest MGSM score is 0.923, achieved by Llama 4 Maverick from Meta.
31 models have been evaluated on the MGSM benchmark, with 0 verified results and 30 self-reported results.
MGSM is categorized under math and reasoning. The benchmark evaluates text models with multilingual support.