SAHARA - Title Generation: leaderboard
Title generation for texts in African languages, scored against reference titles, from the SAHARA African-language LLM leaderboard (UBC-NLP, ACL 2025). The task score is the mean over the languages the task covers; spBLEU-1K on a 0-100 scale, higher is better.
Metric: spBLEU-1K. Source: huggingface.co. Saturation forecast: Around 2030. 48 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemini 3 Pro (Preview) | 17.06 | #71 |
| 2 | Gemini 2.5 Pro | 14.58 | #159 |
| 3 | Claude Sonnet 4 (20250514) | 13.62 | #223 |
| 4 | GPT-5.6 Luna | 13.24 | #79 |
| 5 | Qwen 3.6 27B | 12.66 | #182 |
| 6 | Qwen 3.5 27B | 11.72 | #192 |
| 7 | Qwen 3.6 35B A3B | 11.53 | #238 |
| 8 | Llama 3.1 70B Instruct | 11.07 | #567 |
| 9 | Llama 3.3 70B Instruct | 10.34 | #541 |
| 10 | Qwen 3.5 9B | 9.73 | #391 |
| 11 | Llama 4 Scout Instruct | 9.72 | #547 |
| 12 | GPT-4.1 | 9.48 | #258 |
| 13 | GPT-5 | 9.19 | #105 |
| 14 | Gemini 2.5 Flash | 9.05 | #255 |
| 15 | Gemma 4 31B (IT) | 8.92 | #178 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=sahara-title-generation · How It Works · Data refreshed daily, snapshot 2026-11-05.