OpenBookQA: leaderboard
Open Book Question Answering: 5,957 elementary science questions that require combining an 'open book' of 1,326 core science facts with broad commonsense knowledge.
Metric: Accuracy (%). Source: huggingface.co. Source: epoch.ai. Saturation forecast: Estimated already saturated. 42 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Median Human | 91.6 | |
| 2 | Phi-3 Mini 4K Instruct | 88 | #1297 |
| 3 | Phi-3-small-8k-instruct | 88 | #967 |
| 4 | Phi-3-medium-128k-instruct | 87.4 | #1013 |
| 5 | GPT-3.5 Turbo (1106) | 86 | #961 |
| 6 | Mixtral 8x7B (v0.1) | 85.8 | #1083 |
| 7 | Llama 3 8B Instruct | 82.6 | #1129 |
| 8 | Mistral-7B-v0.1 | 79.8 | #1326 |
| 9 | gemma-7B | 78.6 | #1384 |
| 10 | Phi-2 | 73.6 | #1511 |
| 11 | text-davinci-001 | 65.4 | #1512 |
| 12 | Llama 2 70B Base | 60.2 | #1025 |
| 13 | LLaMA-65B | 60.2 | #1303 |
| 14 | Llama 2 7B | 58.6 | #1520 |
| 15 | LLaMA-7B | 57.2 | #1564 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=openbookqa · How It Works · Data refreshed daily, snapshot 2026-10-20.