# MGSM Bengali (Kaggle v1) (MGSM Bengali (Kaggle)) > Tests mathematical reasoning on the Bengali translation of 250 grade-school math problems from GSM8K. Canonical page: https://benchlm.ai/benchmarks/mgsm-bengali - Category: [Mathematics](/math) - Last updated: 2026-10-20 source capture These results cover the Bengali translation of 250 grade-school math questions. The public Kaggle download includes configurations hidden by the website's default model filters. We retain the published evaluation dates where available. The capture date is when the source was downloaded. Prompting settings remain unknown, and these results do not enter overall or category rankings. These are published source configurations. Compare the task version and disclosed settings before treating a score difference as a model difference. - [Kaggle leaderboard and Apache 2.0 declaration](https://www.kaggle.com/benchmarks/open-benchmarks/mgsm-bengali) - [Published leaderboard download](https://www.kaggle.com/api/v1/benchmarks/open-benchmarks/mgsm-bengali/leaderboard) - [MGSM paper](https://arxiv.org/abs/2210.03057) - [Apache 2.0 license copy](https://benchlm.ai/licenses/kaggle-benchmarks-Apache-2.0.txt) ## About MGSM Bengali (Kaggle) - Year: 2022 - Tasks: 250 Bengali math problems - Format: Problem accuracy (%) - Paper: [Language Models are Multilingual Chain-of-Thought Reasoners](https://arxiv.org/abs/2210.03057) The Bengali subset has 250 manually translated problems. We show the exact published Kaggle Score and keep its source model profiles and evaluation dates. This single-language protocol is separate from the multilingual MGSM aggregate and the Vals evaluation. MGSM Bengali (Kaggle) is currently displayed on BenchLM for reference, but it is excluded from the weighted scoring formula. ## Leaderboard (46 models) | Rank | Model | Configuration | Creator | Score | |------|-------|---------------|---------|-------| | 1 | [Claude Opus 4.8](https://www.kaggle.com/benchmarks/open-benchmarks/mgsm-bengali) | Evaluation date not reported | — | 95.6% | | 2 | [Claude Opus 4](https://www.kaggle.com/benchmarks/open-benchmarks/mgsm-bengali) | Evaluated 2025-07-01 | — | 94.4% | | 3 | [GPT-5](https://www.kaggle.com/benchmarks/open-benchmarks/mgsm-bengali) | Evaluated 2025-08-19 | — | 94.0% | | 4 | [Gemini 2.5 Pro Preview](https://www.kaggle.com/benchmarks/open-benchmarks/mgsm-bengali) | Evaluated 2025-06-30 | — | 92.8% | | 5 | [Claude Opus 4.1](https://www.kaggle.com/benchmarks/open-benchmarks/mgsm-bengali) | Evaluated 2025-08-20 | — | 92.8% | | 6 | [Gemma 4 26B A4B](https://www.kaggle.com/benchmarks/open-benchmarks/mgsm-bengali) | Evaluation date not reported | — | 92.4% | | 7 | [o4 mini](https://www.kaggle.com/benchmarks/open-benchmarks/mgsm-bengali) | Evaluated 2025-06-30 | — | 91.6% | | 8 | [gpt-oss-120b](https://www.kaggle.com/benchmarks/open-benchmarks/mgsm-bengali) | Evaluated 2025-08-19 | — | 91.2% | | 9 | [gpt-oss-20b](https://www.kaggle.com/benchmarks/open-benchmarks/mgsm-bengali) | Evaluated 2025-08-19 | — | 91.2% | | 10 | [Claude 3.7 Sonnet](https://www.kaggle.com/benchmarks/open-benchmarks/mgsm-bengali) | Evaluated 2025-07-01 | — | 90.4% | | 11 | [DeepSeek-V3](https://www.kaggle.com/benchmarks/open-benchmarks/mgsm-bengali) | Evaluated 2025-06-30 | — | 89.2% | | 12 | [deepseek-v3.1](https://www.kaggle.com/benchmarks/open-benchmarks/mgsm-bengali) | Evaluated 2025-08-27 | — | 89.2% | | 13 | [Gemini 1.5 Pro](https://www.kaggle.com/benchmarks/open-benchmarks/mgsm-bengali) | Evaluated 2025-06-30 | — | 88.8% | | 14 | [GPT-4o](https://www.kaggle.com/benchmarks/open-benchmarks/mgsm-bengali) | Evaluated 2025-06-30 | — | 88.8% | | 15 | [DeepSeek-R1](https://www.kaggle.com/benchmarks/open-benchmarks/mgsm-bengali) | Evaluated 2025-06-30 | — | 88.8% | | 16 | [Claude 3.5 Sonnet](https://www.kaggle.com/benchmarks/open-benchmarks/mgsm-bengali) | Evaluated 2025-06-30 | — | 88.4% | | 17 | [o3 mini](https://www.kaggle.com/benchmarks/open-benchmarks/mgsm-bengali) | Evaluated 2025-06-30 | — | 88.4% | | 18 | [Grok 3 Mini](https://www.kaggle.com/benchmarks/open-benchmarks/mgsm-bengali) | Evaluated 2025-06-30 | — | 88.4% | | 19 | [Gemini 3 Pro Preview](https://www.kaggle.com/benchmarks/open-benchmarks/mgsm-bengali) | Evaluated 2025-11-14 | — | 88.4% | | 20 | [Gemini 1.5 Flash](https://www.kaggle.com/benchmarks/open-benchmarks/mgsm-bengali) | Evaluated 2025-06-30 | — | 87.6% | | 21 | [Gemini 2.5 Flash Preview](https://www.kaggle.com/benchmarks/open-benchmarks/mgsm-bengali) | Evaluated 2025-06-30 | — | 87.6% | | 22 | [o3](https://www.kaggle.com/benchmarks/open-benchmarks/mgsm-bengali) | Evaluated 2025-07-01 | — | 87.6% | | 23 | [Claude Sonnet 4](https://www.kaggle.com/benchmarks/open-benchmarks/mgsm-bengali) | Evaluated 2025-07-01 | — | 85.6% | | 24 | [Gemini 3.5 Flash](https://www.kaggle.com/benchmarks/open-benchmarks/mgsm-bengali) | Evaluation date not reported | — | 83.5% | | 25 | [Mistral Large 2](https://www.kaggle.com/benchmarks/open-benchmarks/mgsm-bengali) | Evaluated 2025-07-01 | — | 83.2% | | 26 | [Grok 4](https://www.kaggle.com/benchmarks/open-benchmarks/mgsm-bengali) | Evaluated 2025-09-02 | — | 82.8% | | 27 | [GPT-4o mini](https://www.kaggle.com/benchmarks/open-benchmarks/mgsm-bengali) | Evaluated 2025-06-30 | — | 82.0% | | 28 | [Claude 3.5 Haiku](https://www.kaggle.com/benchmarks/open-benchmarks/mgsm-bengali) | Evaluated 2025-07-01 | — | 78.8% | | 29 | [Gemini 1.5 Flash 8B](https://www.kaggle.com/benchmarks/open-benchmarks/mgsm-bengali) | Evaluated 2025-06-30 | — | 78.4% | | 30 | [Grok 3](https://www.kaggle.com/benchmarks/open-benchmarks/mgsm-bengali) | Evaluated 2025-06-30 | — | 78.0% | | 31 | [Gemma 3 12B](https://www.kaggle.com/benchmarks/open-benchmarks/mgsm-bengali) | Evaluated 2025-06-30 | — | 74.4% | | 32 | [Codestral](https://www.kaggle.com/benchmarks/open-benchmarks/mgsm-bengali) | Evaluated 2025-07-01 | — | 74.0% | | 33 | [Gemini 2.0 Flash](https://www.kaggle.com/benchmarks/open-benchmarks/mgsm-bengali) | Evaluated 2025-06-30 | — | 72.8% | | 34 | [o1](https://www.kaggle.com/benchmarks/open-benchmarks/mgsm-bengali) | Evaluated 2025-06-30 | — | 66.8% | | 35 | [Qwen 3 235B A22B Thinking](https://www.kaggle.com/benchmarks/open-benchmarks/mgsm-bengali) | Evaluated 2025-08-19 | — | 65.6% | | 36 | [Gemma 3 4B](https://www.kaggle.com/benchmarks/open-benchmarks/mgsm-bengali) | Evaluated 2025-06-30 | — | 64.8% | | 37 | [Mixtral 8x22B](https://www.kaggle.com/benchmarks/open-benchmarks/mgsm-bengali) | Evaluated 2025-07-01 | — | 58.8% | | 38 | [Gemma 4 31B](https://www.kaggle.com/benchmarks/open-benchmarks/mgsm-bengali) | Evaluation date not reported | — | 57.2% | | 39 | [Ministral 8B](https://www.kaggle.com/benchmarks/open-benchmarks/mgsm-bengali) | Evaluated 2025-07-01 | — | 56.0% | | 40 | [GPT-4.1](https://www.kaggle.com/benchmarks/open-benchmarks/mgsm-bengali) | Evaluated 2025-06-30 | — | 54.0% | | 41 | [Granite 4.0 Small](https://www.kaggle.com/benchmarks/open-benchmarks/mgsm-bengali) | Evaluated 2025-09-30 | — | 53.2% | | 42 | [Grok 2](https://www.kaggle.com/benchmarks/open-benchmarks/mgsm-bengali) | Evaluated 2025-06-30 | — | 50.4% | | 43 | [Gemma 3 27B](https://www.kaggle.com/benchmarks/open-benchmarks/mgsm-bengali) | Evaluated 2025-06-30 | — | 48.8% | | 44 | [o1 mini](https://www.kaggle.com/benchmarks/open-benchmarks/mgsm-bengali) | Evaluated 2025-06-30 | — | 46.4% | | 45 | [GPT-3.5 Turbo](https://www.kaggle.com/benchmarks/open-benchmarks/mgsm-bengali) | Evaluated 2025-06-30 | — | 32.8% | | 46 | [Ministral 3B](https://www.kaggle.com/benchmarks/open-benchmarks/mgsm-bengali) | Evaluated 2025-07-01 | — | 24.8% | ## FAQ ### What does MGSM Bengali (Kaggle) measure? Tests mathematical reasoning on the Bengali translation of 250 grade-school math problems from GSM8K. ### Which source row has the highest reported MGSM Bengali (Kaggle) value? Claude Opus 4.8 has the highest reported MGSM Bengali (Kaggle) value at 95.6%. These are published source configurations. Compare the task version and disclosed settings before treating a score difference as a model difference. ### How many models are evaluated on MGSM Bengali (Kaggle)? The 2026-10-20 source capture contains 46 AI models. ### Does MGSM Bengali (Kaggle) affect BenchLM's overall score? Not directly. MGSM Bengali (Kaggle) is still displayed on BenchLM for reference, but it is excluded from the weighted scoring formula.