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: epoch.ai. Status: saturated. 42 models tracked.

Top models

#ModelScore
1Phi-3 Mini 4K Instruct88
2Phi-3-small-8k-instruct88
3Phi-3-medium-128k-instruct87.4
4GPT-3.5 Turbo (1106)86
5Mixtral 8x7B (v0.1)85.8
6Median Human85
7Llama 3 8B Instruct82.6
8Mistral-7B-v0.179.8
9gemma-7B78.6
10Phi-273.6
11text-davinci-00165.4
12falcon-180B64.2
13Llama 2 70B Base60.2
14LLaMA-65B60.2
15Llama 2 7B58.6

Interactive version: theaggregate.ai/benchmark?slug=openbookqa · How the rankings work · Data refreshed daily, snapshot 2026-07-22.