SAHARA - Overall: leaderboard
Sahara Score from the SAHARA African-language LLM leaderboard (UBC-NLP, ACL 2025): mean of four task-cluster scores covering text classification, text generation, multiple-choice/comprehension/reasoning and token-level tasks across African languages. Scale 0-100, higher is better.
Metric: Sahara Score (mean of cluster scores). Source: huggingface.co. Saturation forecast: Around April 2028. 48 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemini 3 Pro (Preview) | 55.67 | #71 |
| 2 | Gemini 2.5 Pro | 51.85 | #159 |
| 3 | GPT-5 | 50.03 | #105 |
| 4 | GPT-5.1 (2025-11-13) | 48.33 | #107 |
| 5 | GPT-5.6 Luna | 47.53 | #79 |
| 6 | Claude Sonnet 4 (20250514) | 40.82 | #223 |
| 7 | Gemini 2.5 Flash | 40.61 | #255 |
| 8 | Claude Sonnet 4 | 37.11 | #207 |
| 9 | GPT-4.1 | 36.04 | #258 |
| 10 | Qwen 3.5 27B | 35.96 | #192 |
| 11 | Qwen 3.6 27B | 35.41 | #182 |
| 12 | Qwen 3.6 35B A3B | 32.09 | #238 |
| 13 | c4ai-command-a-03-2025 | 29.93 | #468 |
| 14 | Llama 4 Scout Instruct | 29.19 | #547 |
| 15 | Gemma 3 27B (IT) | 28.07 | #528 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=sahara-overall · How It Works · Data refreshed daily, snapshot 2026-11-05.