Home/Comparar/Meta Llama 3 8B Instruct vs Qwen2 7B Instruct

Meta Llama 3 8B Instruct vs Qwen2 7B Instruct

Lado a lado. Cada métrica. Cada benchmark.

Pontuações de benchmark

25 benchmarks · Meta Llama 3 8B Instruct: 3, Qwen2 7B Instruct: 22

Benchmark	Categoria	Meta Llama 3 8B Instruct	Qwen2 7B Instruct
BBH (HuggingFace)	general	28.2	37.8
GPQA	knowledge	1.2	6.4
IFEval	language	74.1	56.8
JCommonsenseQA	language	87.7	89.1
JMMLU	language	46.7	56.5
JNLI	language	61.1	81.3
JSQuAD	language	89.5	89.6
LLM-JP — Overall	language	49.6	51.7
MATH Level 5	math	8.7	27.6
MMLU-PRO	knowledge	29.6	31.6
MMMLU — Arabic	language	40.5	50.7
MMMLU — Bengali	language	36.4	43.4
MMMLU — Chinese	language	51.4	61.8
MMMLU — French	language	55.8	60.8
MMMLU — German	language	53.5	57.1
MMMLU — Hindi	language	41.4	45.1
MMMLU — Indonesian	language	51.0	54.1
MMMLU — Italian	language	53.3	59.0
MMMLU — Japanese	language	42.3	56.6
MMMLU — Korean	language	46.5	54.0
MMMLU — Portuguese	language	55.5	60.1
MMMLU — Spanish	language	55.8	60.2
MMMLU — Swahili	language	37.5	34.3
MMMLU — Yoruba	language	31.0	30.2
MUSR	reasoning	1.6	7.4