GitLab Transcend - Intelligent orchestration
Mini toolsMGSM
Multilingual Grade School Math
GSM8K translated into 10 typologically diverse languages. Tests cross-lingual mathematical reasoning.
Language
Text
accuracy
Max 100.0%
Released Oct 2022
5
Results
5
Models scored
91.1%
Top: Llama 3.3
90.5%
Median
Best results
Top primary scores; one row per model.
Frontier over time
Each dot is one model result; the line traces the running best score.
All results
Showing all configurations including non-primary alternates. · Show only primary
| # | Model | Score | Conditions | Eval date | Source | Flags |
|---|---|---|---|---|---|---|
| 1 | Claude Sonnet 3.5 | 91.6% | 0-shot · CoT · standard | Jun 20, 2024 | self reported | |
| 2 | Llama 3.3 | 91.1% | 0-shot | Dec 6, 2024 | self reported | primary |
| 3 | Claude Opus 3 | 90.7% | 0-shot | Oct 22, 2024 | self reported | primary |
| 4 | GPT-4o | 90.5% | — | Apr 16, 2025 | self reported | primary |
| 5 | Gemini 1.5 Pro | 87.5% | 8-shot · standard | May 1, 2024 | self reported | |
| 6 | Nemotron 3 Super | 87.5% | 8-shot | Apr 3, 2026 | self reported | primary |
| 7 | Gemini 1.5 Flash | 82.6% | 8-shot · standard | May 1, 2024 | self reported | |
| 8 | Gemini Ultra | 79.0% | 8-shot · CoT | Dec 6, 2023 | paper | |
| 9 | Claude Haiku 3 | 75.1% | 0-shot · CoT · standard | Mar 4, 2024 | self reported | |
| 10 | Llama 3.2 | 58.2% | 0-shot · CoT | Oct 25, 2024 | self reported | primary |
