AfroBench: leaderboard
64 African languages across 15 NLP tasks and 22 datasets (understanding, text generation, knowledge and question answering, math reasoning), from McGill NLP and Masakhane (2025); average score.
Metric: Average Score (%). Source: mcgill-nlp.github.io. Source: huggingface.co. Source: benchmarklist.com. Saturation forecast: Around January 2027. 24 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-5 | 77.7 | #98 |
| 2 | Gemini 3 Pro | 76 | #80 |
| 3 | Gemini 2.5 Pro | 74.5 | #152 |
| 4 | Claude Sonnet 4.5 | 70.6 | #142 |
| 5 | Claude Sonnet 4 | 68.7 | #199 |
| 6 | Gemini 2.5 Flash | 66.7 | #245 |
| 7 | Gemini 2.0 Flash | 66.4 | #342 |
| 8 | GPT-4o (2024-08-06) | 65.8 | #336 |
| 9 | GPT-4.1 | 65.7 | #248 |
| 10 | Gemini 1.5 Pro | 62.8 | #445 |
| 11 | Claude 3.7 Sonnet | 60.3 | #249 |
| 12 | Gemma 3 27B | 51 | #607 |
| 13 | Gemma 2 27B | 44.9 | #806 |
| 14 | Gemma 2 9B | 40.8 | #958 |
| 15 | Llama 3.1 70B | 40 | #592 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=afrobench · How It Works · Data refreshed daily, snapshot 2026-10-20.