MGSM Bengali
Math · 2022-10-06
Bengali language split of the MGSM multilingual grade-school math benchmark.
Top models (higher is better)
| Model | Score |
|---|---|
| Opus 4.5 | 96.8 |
| Sonnet 4.5 | 94.8 |
| GPT-5.2 | 94.8 |
| GPT-5 | 94.8 |
| GPT-5 Mini | 94.0 |
| Opus 4 | 93.6 |
| Opus 4.1 | 93.2 |
| GPT-5.1 | 93.2 |
| gpt-oss-120b | 92.8 |
| O4 Mini | 92.8 |
| Haiku 4.5 | 92.4 |
| Llama 4 Maverick Instruct | 92.4 |
| DeepSeek-V3 | 91.6 |
| Sonnet 4 | 91.2 |
| Qwen3 Max Preview | 91.2 |
| Claude 3.5 Sonnet (Oct 2024) | 90.4 |
| Claude 3.7 Sonnet | 90.4 |
| DeepSeek-V3-0324 | 90.4 |
| Gemini 1.5 Pro 002 | 90.4 |
| Gemini 3 Pro Preview | 90.4 |
| GPT-4o (2024-11-20) | 90.4 |
| o3-mini | 90.0 |
| GLM-4.5 | 89.6 |
| GPT-4o (2024-08-06) | 89.6 |
| Qwen3 Max (rolling alias) | 89.6 |
| Mistral Medium 3 | 89.2 |
| Grok 3 Mini | 88.8 |
| DeepSeek-R1 | 88.4 |
| Qwen3-235B-A22B | 88.4 |
| Llama 3.3 70B Instruct | 88.0 |
| Grok 4 | 87.6 |
| Grok 4.1 Fast | 87.6 |
| Kimi K2 Instruct | 87.6 |
| Mistral Large 2.1 (Instruct 2411) | 86.8 |
| Grok 4 Fast | 86.4 |
| Kimi K2 Thinking | 86.4 |
| GPT-5 Nano | 86.0 |
| o3 | 86.0 |
| Gemini 1.5 Flash 002 | 85.6 |
| Command A | 84.4 |