AfriScience-MT - Document Level (Zero-Shot): leaderboard
Whole-document translation of AfriScience-MT scientific summaries (Abdulmumin et al., 2026) between English and Amharic, Hausa, Luganda, Northern Sotho, Yoruba and isiZulu, with a plain prompt and no examples. SSA-COMET on a 0-100 scale averaged over the 12 directions of the test split; higher is better.
Metric: SSA-COMET (0-100; whole-summary translation, mean over the 12 English<->African directions of the test split; vanilla 0-shot prompt, temperature 0). Source: arxiv.org. Saturation forecast: Around 2031. 11 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-5.4 | 48.3 | #121 |
| 2 | Gemini 3.1 Flash Lite | 48.3 | #363 |
| 3 | Tiny Aya Global | 40.6 | #6489 |
| 4 | Llama 3.1 8B Instruct | 36.8 | #4277 |
| 5 | Gemini 1.5 Flash | 34.5 | #1671 |
| 6 | Gemma 2 9B (IT) | 34.3 | #2275 |
| 7 | GPT-4o | 32.1 | #741 |
No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #5 Claude Opus 5.5.
Interactive version: theaggregate.ai/benchmark?slug=afriscience-mt-document-level-zero-shot · How It Works · Data refreshed daily, snapshot 2026-11-12.