SAHARA - News Classification: leaderboard
News classification, assigning news texts in African languages to categories, from the SAHARA African-language LLM leaderboard (UBC-NLP, ACL 2025). The task score is the mean over the languages the task covers; Macro-F1 on a 0-100 scale, higher is better.
Metric: Macro-F1 (%). Source: huggingface.co. Saturation forecast: Around June 2027. 48 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Claude Sonnet 4 (20250514) | 40.09 | #223 |
| 2 | Claude Sonnet 4 | 38.17 | #207 |
| 3 | GPT-5 | 36.99 | #105 |
| 4 | Gemma 3 27B (IT) | 35.56 | #528 |
| 5 | Gemini 3 Pro (Preview) | 35.51 | #71 |
| 6 | Llama 3.1 70B Instruct | 35.26 | #567 |
| 7 | Qwen 3.6 27B | 35.1 | #182 |
| 8 | DeepSeek R1 Distill Llama 70B | 34.81 | #614 |
| 9 | Qwen 3.6 35B A3B | 34.66 | #238 |
| 10 | GPT-5.1 (2025-11-13) | 34.57 | #107 |
| 11 | Gemini 2.5 Pro | 33.77 | #159 |
| 12 | Llama 3.3 70B Instruct | 33.31 | #541 |
| 13 | Qwen 3.5 27B | 33.05 | #192 |
| 14 | GPT-5.6 Luna | 32.98 | #79 |
| 15 | Gemini 2.5 Flash | 32.55 | #255 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=sahara-news-classification · How It Works · Data refreshed daily, snapshot 2026-11-05.