Benchmark Leaderboard

Top models by HellaSwag score

#	Model	Score	Version	Source
1	GPT-4o (05-13)	96.4	10-shot	HELM, Open LLM Leaderboard
2	Claude 3.5 Sonnet	96.2	10-shot	HELM, official documentation
3	Llama 3.1 405B	95.8	10-shot	Open LLM Leaderboard, Meta official
4	DeepSeek V3	95.7	10-shot	Open LLM Leaderboard, DeepSeek official
5	Qwen2.5 72B	95.6	10-shot	research
6	Qwen2.5 72B Instruct	95.6	standard	Open LLM Leaderboard
7	Llama 3.1 70B Instruct	94.2	10-shot	Open LLM Leaderboard
8	Mistral Medium	93.9	10-shot	research
9	Mistral Large 2	93.8	10-shot	Open LLM Leaderboard
10	Mixtral 8x22B v0.1	93.8	10-shot	research
11	Falcon 180B	92.7	10-shot	Open LLM Leaderboard
12	Gemma 2 27B	92.6	10-shot	research
13	Llama 3 70B	92.4	10-shot	research
14	Qwen2 7B	92	10-shot	research
15	Mistral NeMo Instruct (2407)	91.8	10-shot	Open LLM Leaderboard
16	StarCoder2 15B	91.7	10-shot	research
17	DeepSeek Coder V2 Lite	91.4	10-shot	Open LLM Leaderboard
18	Llama 3 8B Instruct	91.1	10-shot	research
19	Mixtral 8x7B	90.9	10-shot	Open LLM Leaderboard
20	Command R	90.8	10-shot	research