Llama 3.1 Instruct 70B
MetaLlama开源权重Llama 3.1 Community License
描述
Llama 3.1 70B Instruct is a large language model optimized for multilingual dialogue use cases. It outperforms many available open source and closed chat models on common industry benchmarks.
发布日期
2024-07-23
参数规模
70.0B
上下文长度
131K
支持模态
text
能力雷达图
25
general
23
coding
20
reasoning
30
science
70
agents
0
multimodal
排行榜排名
基准测试分数 (LLM Stats)
(LLM Stats (zeroeval))Chat
IFEvalGoogle Research (2023)
87.5%自报
General
BFCL
84.8%自报
MMLU
83.6%自报
Multipl-E MBPP
62.0%自报
Nexus
56.7%自报
Language
MMLU (CoT)
86.0%自报
MMLU-Pro
66.4%自报
Multipl-E HumanEval
65.5%自报
Math
GSM-8K (CoT)
95.1%自报
Multilingual MGSM (CoT)
86.9%自报
MATH (CoT)
68.0%自报
Reasoning
ARC-C
94.8%自报
API-Bank
90.0%自报
MBPP ++ base version
86.0%自报
HumanEvalOpenAI (2021)
80.5%自报
DROP
79.6%自报
GPQANYU + Cohere + Anthropic (2023)
41.7%自报
Gorilla Benchmark API Bench
29.7%自报
AA 评测指数
(Artificial Analysis)Mmlu Pro(TIGER-Lab (Univ. of Waterloo, Toronto, CMU, 2024))67.6
Math 500(OpenAI (2024), subset of Hendrycks et al. MATH (2021))64.9
Gpqa(NYU + Cohere + Anthropic (2023))40.9
Ifbench(Google Research (2023))34.4
Livecodebench(UC Berkeley + MIT + Cornell (2024))23.2
Aime(MAA (Mathematical Association of America))17.3
Tau2(Sierra + U Toronto + Vector Institute (2025))15.2
Intelligence Index(Artificial Analysis)6.6
Hle(Center for AI Safety + Scale AI (2025))4.5
Math Index(Artificial Analysis)4.0
Aime 25(MAA (Mathematical Association of America))4.0
Terminalbench Hard(Stanford × Laude Institute (2026))3.0
LLM Stats 分类评分
(LLM Stats (zeroeval))Chat90
Instruction Following90
Structured Output90
Language80
Legal80
Math80
Finance80
Healthcare80
Reasoning70
General70
Tool Calling70
Code60
Physics40
Biology40
Chemistry40
定价
输入价格$0.56 / 1M tokens
输出价格$0.56 / 1M tokens
混合价格(3:1)$0.56 / 1M tokens
速度
Tokens/秒0.0
首Token延迟0.00s
首回答延迟0.00s
供应商价格排行
供应商价格排行
4 个供应商
最便宜: DeepInfra最贵: Meta
供应商输入输出
1DeepInfra最便宜
$0
$0
2OpenRouter
$0.4
$0.4
3Kilo Gateway
$0.4
$0.4
4Meta主要
$0.56
$0.56
比较该模型在不同 API 供应商之间的定价。