SAHARA - Language Identification: leaderboard
Language identification (lid), naming which African language a text is written in, 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. Most models score near zero.
Metric: Macro-F1 (%). Source: huggingface.co. Saturation forecast: Around 2031. 48 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemini 3 Pro (Preview) | 11.84 | #71 |
| 2 | Gemma 2 27B (IT) | 4.61 | #652 |
| 3 | c4ai-command-a-03-2025 | 4.57 | #468 |
| 4 | GPT-4.1 | 4.48 | #258 |
| 5 | GPT-5 | 4.23 | #105 |
| 6 | GPT-5.1 (2025-11-13) | 4.15 | #107 |
| 7 | Qwen 3.5 27B | 4.06 | #192 |
| 8 | Qwen 3.5 4B | 4.02 | #514 |
| 9 | Gemma 3 27B (IT) | 4 | #528 |
| 10 | Claude Sonnet 4 (20250514) | 3.73 | #223 |
| 11 | Qwen 3.6 35B A3B | 3.7 | #238 |
| 12 | Gemma 2 9B (IT) | 3.6 | #795 |
| 13 | GPT-5.6 Luna | 3.54 | #79 |
| 14 | aya-23-35B | 3.34 | #873 |
| 15 | Command-R+ (08-2024) | 3.32 | #838 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=sahara-language-identification · How It Works · Data refreshed daily, snapshot 2026-11-05.