测试版
排行榜/Meta Llama 3 8B Instruct
Meta logo

Meta Llama 3 8B Instruct

开源

来自 Meta · 发布于 2024-04-17

45.2
平均分
N/A
输入价格
N/A
输出价格
N/A
上下文窗口
text-generation
类型

Tested on 25 benchmarks with 45.2% average. Top scores: JSQuAD (89.5%), JCommonsenseQA (87.7%), IFEval (74.1%).

基准测试类别分数Bar
JSQuADlanguage89.5
JCommonsenseQAlanguage87.7
IFEvallanguage74.1
JNLIlanguage61.1
MMMLU — Spanishlanguage55.8
MMMLU — Frenchlanguage55.8
MMMLU — Portugueselanguage55.5
MMMLU — Germanlanguage53.5
MMMLU — Italianlanguage53.3
MMMLU — Chineselanguage51.4
MMMLU — Indonesianlanguage51.0
LLM-JP — Overalllanguage49.6
JMMLUlanguage46.7
MMMLU — Koreanlanguage46.5
MMMLU — Japaneselanguage42.3
MMMLU — Hindilanguage41.4
MMMLU — Arabiclanguage40.5
MMMLU — Swahililanguage37.5
MMMLU — Bengalilanguage36.4
MMMLU — Yorubalanguage31.0
MMLU-PROknowledge29.6
BBH (HuggingFace)general28.2
MATH Level 5math8.7
MUSRreasoning1.6
GPQAknowledge1.2