HELM Arabic - Alghafa - Multiple Choice Sentiment Task: leaderboard
HELM Arabic (release v2.2.0) AlGhafa Multiple Choice Sentiment Task task: sentiment classification of Arabic texts, from the native AlGhafa Arabic multiple-choice benchmark; 1,000 items, exact match. Percent correct, 0-100, higher is better.
Metric: Exact Match (%). Source: crfm.stanford.edu. Saturation forecast: Around 2033. 41 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Claude Haiku 4.5 (20251001) | 46.1 | #264 |
| 2 | Qwen 3.5 397B A17B | 45.5 | #154 |
| 3 | GPT-4.1 (2025-04-14) | 43.9 | #257 |
| 4 | Llama 4 Scout Instruct | 43.7 | #547 |
| 5 | DeepSeek V3.1 | 43.7 | #275 |
| 6 | Gemini 2.5 Flash (Non-reasoning) | 43.7 | #255 (Gemini 2.5 Flash) |
| 7 | Llama 3.3 70B Instruct | 43.3 | #541 |
| 8 | GPT-5.4 (2026-03-05) | 43.1 | #63 |
| 9 | Qwen 2.5 72B Instruct | 43 | #469 |
| 10 | Claude Opus 4.7 | 43 | #51 |
| 11 | Llama 4 Maverick Instruct FP8 | 42.7 | #424 |
| 12 | GPT-5.1 (2025-11-13) | 42.4 | #107 |
| 13 | Gemini 2.5 Flash Lite (Non-reasoning) | 42.2 | #432 (Gemini 2.5 Flash Lite) |
| 14 | Claude Sonnet 4.6 | 42.1 | #94 |
| 15 | AceGPT-v2-32B-Chat | 42.1 | #518 |
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-alghafa-multiple-choice-sentiment-task · How It Works · Data refreshed daily, snapshot 2026-11-05.