HELM Arabic - Arabic Exams - Biology: leaderboard
HELM Arabic (release v2.2.0) Arabic EXAMS, Biology questions: Arabic high-school examination multiple-choice questions from the Arabic subset of EXAMS (OALL mirror of the AceGPT version); 35 questions, exact match. Percent correct, 0-100, higher is better.
Metric: Exact Match (%). Source: crfm.stanford.edu. Saturation forecast: Around 2034. 41 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Llama 3.3 70B Instruct | 45.71 | #541 |
| 2 | GPT-4.1 Mini | 45.71 | #363 |
| 3 | Qwen 3 Next 80B A3B Instruct | 45.71 | #381 |
| 4 | Claude Sonnet 4.6 | 42.86 | #94 |
| 5 | GPT-5.4 (2026-03-05) | 42.86 | #63 |
| 6 | Claude Opus 4.7 | 40 | #51 |
| 7 | Qwen 3.5 397B A17B | 40 | #154 |
| 8 | Gemma 4 31B (IT) | 40 | #178 |
| 9 | Gemini 2.5 Flash (Non-reasoning) | 40 | #255 (Gemini 2.5 Flash) |
| 10 | jais-adapted-13B-chat | 40 | #948 |
| 11 | Falcon3-10B-Instruct | 40 | #1147 |
| 12 | GPT-5.4 Mini (2026-03-17) | 40 | #180 |
| 13 | Qwen 3.5 9B | 37.14 | #391 |
| 14 | GPT-4.1 Nano | 37.14 | #730 |
| 15 | Llama 4 Scout Instruct | 37.14 | #547 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=helm-arabic-arabic-exams-biology · How It Works · Data refreshed daily, snapshot 2026-11-05.