OpenBookQA

Models scored

70

Best score

0.88

Metric

Accuracy

OpenBookQA is an AI evaluation tracked by Epoch AI, with 70 scored model versions on record. The scores here are reported by third parties rather than produced by Epoch AI running the evaluation itself.

Scored models span November 2019 to April 2024. The best score recorded is 0.88, measured as Accuracy. Models come from Cerebras Systems, Databricks, Technology Innovation Institute, Google DeepMind, Google, OpenAI, EleutherAI,LAION, EleutherAI and others.

Full record

Score metric: Accuracy
Models scored: 70
Best score recorded: 0.88
Earliest model scored: Nov 5, 2019
Most recent model scored: Apr 23, 2024
Third-party reported: Yes

Leaderboard

01	Phi-3-small-8k-instruct	Microsoft	0.88
02	Phi-3-mini-4k-instruct	Microsoft	0.88
03	Phi-3-medium-128k-instruct	Microsoft	0.87
04	gpt-3.5-turbo-1106	OpenAI	0.86
05	Mixtral-8x7B-v0.1	Mistral AI	0.86
06	Meta-Llama-3-8B-Instruct	Meta AI	0.83
07	Mistral-7B-v0.1	Mistral AI	0.8
08	gemma-7b	Google DeepMind	0.79
09	phi-2	Microsoft	0.74
10	PaLM 540B	Google Research	0.68
11	text-davinci-001	OpenAI	0.65
12	falcon-180B	Technology Innovation Institute	0.64
13	GLaM (MoE)	Google	0.63
14	LLaMA-65B	Meta AI	0.6
15	Llama-2-70b-hf	Meta AI	0.6
16	LLaMA-33B	Meta AI	0.59
17	Llama-2-7b	Meta AI	0.59
18	Llama-2-34b	Meta AI	0.58
19	PaLM 2-M		0.57
20	LLaMA-7B	Meta AI	0.57
21	Llama-2-13b	Meta AI	0.57
22	falcon-40b	Technology Innovation Institute	0.57
23	LLaMA-13B	Meta AI	0.56
24	PaLM 2-S		0.56
25	mpt-30b	MosaicML	0.52
26	falcon-7b	Technology Innovation Institute	0.52
27	mpt-7b	MosaicML	0.51
28	PaLM 62B		0.5
29	xgen-7b-8k-base	Salesforce	0.4
30	RedPajama-INCITE-7B-Base		0.4
31	opt-13b		0.4
32	dolly-v2-12b	Databricks	0.39
33	open_llama_7b		0.39
34	gpt-neox-20b	EleutherAI	0.39
35	gpt-j-6b	EleutherAI,LAION	0.38
36	phi-1_5	Microsoft	0.37

More benchmarks

Epoch Capabilities Index MMLU Terminal-Bench GPQA Diamond OTIS Mock AIME 2024-2025 GSM8K ARC-AGI ARC-AGI-2

SourceEpoch AI, 'AI Benchmarking Hub'. Published online at epoch.ai. Retrieved 2026-07-29 from https://epoch.ai/benchmarks. Licensed under CC BY 4.0.

← All benchmarks