MGSM Japanese
Math · 2022-10-06
Japanese language split of the MGSM multilingual grade-school math benchmark.
Top models (higher is better)
| Model | Score |
|---|---|
| Opus 4.5 | 93.6 |
| Gemini 3 Pro Preview | 93.6 |
| GPT-5.1 | 93.2 |
| GPT-5.2 | 92.4 |
| Opus 4.1 | 92.0 |
| Gemini 3 Flash Preview | 92.0 |
| GPT-5 | 92.0 |
| DeepSeek-R1 | 91.6 |
| o3 | 91.6 |
| Sonnet 4.5 | 91.2 |
| Haiku 4.5 | 90.8 |
| Opus 4 | 90.8 |
| DeepSeek-V3 | 90.8 |
| Qwen3-235B-A22B | 90.8 |
| Mistral Medium 3 | 90.4 |
| o3-mini | 90.4 |
| Qwen3 Max (rolling alias) | 90.4 |
| Sonnet 4 | 90.0 |
| DeepSeek-V3-0324 | 90.0 |
| GLM-4.5 | 90.0 |
| MiniMax M2.1 | 90.0 |
| Mistral Large 3 675B Instruct 2512 | 90.0 |
| O1 | 90.0 |
| O4 Mini | 90.0 |
| Claude 3.5 Sonnet (Oct 2024) | 89.6 |
| Claude 3.7 Sonnet | 89.6 |
| DeepSeek-V3.2 | 89.6 |
| Gemini 2.5 Flash Preview (09-2025) | 89.6 |
| GLM 4.6 | 89.6 |
| GLM-4.7 | 89.6 |
| Kimi K2 Instruct | 89.6 |
| GPT-5 Mini | 89.2 |
| gpt-oss-20b | 89.2 |
| Grok 4 Fast | 89.2 |
| Qwen3 Max Preview | 89.2 |
| Gemini 2.5 Flash-Lite Preview (09-2025) | 88.8 |
| GPT-4.1 Mini | 88.8 |
| gpt-oss-120b | 88.8 |
| Grok 4 | 88.8 |
| Grok 2 | 88.0 |