HELM AFR - Winogrande: leaderboard
The "Winogrande" subtask of HELM AFR. Stanford HELM's African-languages project (mmlu-winogrande-afr): MMLU subject exams and Winogrande coreference translated into sixteen African languages, scored against the same items in English so a model's loss from the language switch...
Metric: EM. Source: crfm.stanford.edu. Saturation forecast: Estimated already saturated. 22 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Claude 3.7 Sonnet (20250219) | 68.45 | #205 |
| 2 | Gemini 2.0 Flash | 68.17 | #345 |
| 3 | GPT-4o (2024-08-06) | 67.03 | #337 |
| 4 | Gemini 2.0 Flash Lite | 62.14 | #455 |
| 5 | GPT-4 Turbo | 62.08 | #410 |
| 6 | DeepSeek V3 | 61.1 | #325 |
| 7 | Llama 3.1 405B Instruct | 58.06 | #464 |
| 8 | Llama 3.1 70B Instruct | 54.41 | #567 |
| 9 | Gemma 2 27B (IT) | 54.13 | #653 |
| 10 | Qwen 2.5 72B Instruct | 53.66 | #463 |
| 11 | Claude 3.5 Haiku (20241022) | 52.86 | #591 |
| 12 | Mixtral 8x7B Instruct | 51.97 | #1171 |
| 13 | GPT-3.5 Turbo (0125) | 51.91 | #998 |
| 14 | Llama 3.3 70B Instruct | 51.46 | #537 |
| 15 | Gemma 2 9B (IT) | 50.67 | #795 |
No result here: #4 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=helm-afr-winogrande · How It Works · Data refreshed daily, snapshot 2026-10-24.