HELM Lite - OpenbookQA: leaderboard
HELM Lite v1.13.0 OpenbookQA run: four-option elementary-science questions that combine a core science fact with broader common knowledge; 500 test questions, 5-shot, exact match. Percent correct, 0-100, higher is better; guessing scores 25.
Metric: Exact Match (%). Source: crfm.stanford.edu. Saturation forecast: Estimated already saturated. 91 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Claude 3.5 Sonnet (20240620) | 97.2 | #308 |
| 2 | GPT-4 Turbo | 97 | #417 |
| 3 | GPT-4o (2024-08-06) | 96.8 | #343 |
| 4 | Claude 3.5 Sonnet (20241022) | 96.6 | #302 |
| 5 | GPT-4o (2024-05-13) | 96.6 | #337 |
| 6 | Qwen 2.5 72B Instruct | 96.2 | #469 |
| 7 | GPT-4 (0613) | 96 | #493 |
| 8 | Nova Pro | 96 | #594 |
| 9 | Claude 3 Opus (20240229) | 95.6 | #428 |
| 10 | DeepSeek V3 | 95.4 | #329 |
| 11 | Qwen 2 72B Instruct | 95.4 | #579 |
| 12 | Gemini 1.5 Pro (002) | 95.2 | #368 |
| 13 | GPT-4 Turbo (Preview) | 95 | #416 |
| 14 | Jamba 1.5 Large | 94.8 | #677 |
| 15 | Gemini 2.0 Flash (Preview) | 94.6 | #376 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=helm-lite-openbookqa · How It Works · Data refreshed daily, snapshot 2026-11-05.