HellaSwag

Models scored

113

Best score

0.95

Metric

Overall accuracy

HellaSwag is an AI evaluation tracked by Epoch AI, with 113 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 December 2024. The best score recorded is 0.95, measured as Overall accuracy. Models come from OpenAI, Baichuan, Hugging Face,BigScience, Cerebras Systems, DeepSeek, Databricks, Technology Innovation Institute, Google DeepMind and others.

Full record

Score metric: Overall accuracy
Models scored: 113
Best score recorded: 0.95
Earliest model scored: Nov 5, 2019
Most recent model scored: Dec 26, 2024
Third-party reported: Yes

Leaderboard

01	gpt-4-0314	OpenAI	0.95
02	gpt-4-32k-0314	OpenAI	0.95
03	Llama-3.1-405B	Meta AI	0.89
04	falcon-180B	Technology Innovation Institute	0.89
05	DeepSeek-V3	DeepSeek	0.89
06	DeepSeek-V2	DeepSeek	0.87
07	PaLM 2-L		0.87
08	Mixtral-8x7B-v0.1	Mistral AI	0.87
09	text-davinci-003		0.85
10	Llama-2-70b-hf	Meta AI	0.85
11	falcon-40b	Technology Innovation Institute	0.85
12	Qwen2.5-72B	Alibaba	0.85
13	LLaMA-65B	Meta AI	0.84
14	StableBeluga2	Stability AI	0.84
15	PaLM 2-M		0.84
16	PaLM 540B	Google Research	0.84
17	Qwen2.5-Coder-32B	Alibaba	0.83
18	falcon-11b	Technology Innovation Institute	0.83
19	LLaMA-33B	Meta AI	0.83
20	Phi-3-medium-128k-instruct	Microsoft	0.82
21	Megatron-Turing NLG 530B	Microsoft,NVIDIA	0.82
22	Nemotron-4 15B	NVIDIA	0.82
23	gemma-7b	Google DeepMind	0.82
24	PaLM 2-S		0.82
25	text-davinci-002	OpenAI	0.81
26	Mistral-7B-v0.1	Mistral AI	0.81
27	Llama-2-13b	Meta AI	0.81
28	Qwen2.5-Coder-14B		0.8
29	text-davinci-001	OpenAI	0.79
30	LLaMA-13B	Meta AI	0.79
31	Gopher (280B)	DeepMind	0.79
32	opt-175b	Meta AI	0.79
33	falcon-7b	Technology Innovation Institute	0.78
34	internlm-20b		0.78
35	davinci	OpenAI	0.78

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