SAHM - MCQ Average: leaderboard
Metric: Mean accuracy (%) over the four multiple-choice tasks (accounting exams, business exams, fatwa MCQ, financial sentiment) of SAHM, an Arabic financial and Sharia-compliance benchmark, exact match after normalizing the chosen option, greedy decoding, mean of three runs; higher is better. Source: arxiv.org. Saturation forecast: Around January 2027. 20 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Gemini 3 Flash (Preview) | 80.52 |
| 2 | Claude Opus 4.5 | 80.27 |
| 3 | Claude Sonnet 4.5 | 80.15 |
| 4 | GPT-5 | 76.86 |
| 5 | GPT-4o | 75.96 |
| 6 | Claude Haiku 4.5 | 75.96 |
| 7 | Qwen 2.5 72B Instruct | 75.1 |
| 8 | Llama 3.1 70B Instruct | 73.65 |
| 9 | GPT-4o Mini | 73.61 |
| 10 | Gemma 3 27B (IT) | 71.94 |
| 11 | Fanar-1-9B-Instruct | 61.66 |
| 12 | ALLaM-7B-Instruct-preview | 61.59 |
| 13 | Qwen 2.5 14B Instruct | 61.51 |
| 14 | Gemma 3 4B (IT) | 60.61 |
| 15 | Llama 3.1 8B Instruct | 60.6 |
Interactive version: theaggregate.ai/benchmark?slug=sahm-mcq-average · How It Works · Data refreshed daily, snapshot 2026-10-07.