Arabic Broad Leaderboard - RAG QA: leaderboard
Arabic question answering over supplied context (retrieval-augmented generation): the 41-question RAG QA category of SILMA's Arabic Broad Benchmark, whose answers are scored 0-10 by rules or an LLM judge. Category average score (0-10), higher is better; each model's latest result file is used.
Metric: Average Score (0-10). Source: huggingface.co. Saturation forecast: Estimated already saturated. 121 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GLM-5.3 | 9.73 | #44 |
| 2 | Gemini 2.5 Pro (Preview 05-06) | 9.71 | #156 |
| 3 | Kimi K3 | 9.54 | #35 |
| 4 | Muse Spark 1.1 | 9.51 | #55 |
| 5 | Gemini 3.1 Pro (Preview) | 9.46 | #64 |
| 6 | GPT-5.6 Pro Sol | 9.46 | #21 |
| 7 | GPT-5.5 | 9.41 | #33 |
| 8 | Gemini 3 Pro (Preview) | 9.41 | #71 |
| 9 | Gemini 2.5 Flash (Preview 04-17) | 9.32 | #246 |
| 10 | Gemini 3.5 Flash | 9.29 | #66 |
| 11 | O3 | 9.22 | #132 |
| 12 | MiniMax-M3 | 9.2 | #151 |
| 13 | GLM-5 | 9.17 | #149 |
| 14 | Qwen3.8 2.4T A95B | 9.05 | #59 |
| 15 | DeepSeek V4.1 Flash | 8.9 | #37 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=arabic-broad-leaderboard-rag-qa · How It Works · Data refreshed daily, snapshot 2026-11-05.