跳轉到主要內容
AITier

Qwen2.5 Instruct 72B

AlibabaQwen開源權重Qwen · 商用許可

描述

Qwen2.5-72B-Instruct is an instruction-tuned 72 billion parameter language model, part of the Qwen2.5 series. It is designed to follow instructions, generate long texts (over 8K tokens), understand structured data (e.g., tables), and generate structured outputs, especially JSON. The model supports multilingual capabilities across over 29 languages.

發布日期
2024-09-19
參數規模
72.7B
上下文長度
131K
支援模態
text

能力雷達圖

26
general
28
coding
29
reasoning
35
science
29
agents
0
multimodal

排行榜排名

領域#排名分數來源
程式碼能力榜513
19.0
AA
通用能力榜422
33.0
AA
科學能力517
25.0
AA

基準測試分數 (LLM Stats)

(LLM Stats (zeroeval))

Chat

MT-Bench0.94 / 100自報
IFEvalGoogle Research (2023)84.1%自報

General

AlignBench81.6%自報
Arena Hard81.2%自報
MultiPL-E75.1%自報

Language

MMLU-Redux86.8%自報
MMLU-Pro71.1%自報

Math

GSM8k95.8%自報
MATH83.1%自報
LiveBench52.3%自報

Reasoning

MBPP0.88 / 100自報
HumanEvalOpenAI (2021)86.6%自報
LiveCodeBench55.5%自報
GPQANYU + Cohere + Anthropic (2023)49.0%自報

AA 評測指數

(Artificial Analysis)
Math 500(OpenAI (2024), subset of Hendrycks et al. MATH (2021))
85.8
Mmlu Pro(TIGER-Lab (Univ. of Waterloo, Toronto, CMU, 2024))
72.0
Gpqa(NYU + Cohere + Anthropic (2023))
49.1
Ifbench(Google Research (2023))
36.9
Tau2(Sierra + U Toronto + Vector Institute (2025))
34.5
Livecodebench(UC Berkeley + MIT + Cornell (2024))
27.6
Aime(MAA (Mathematical Association of America))
16.0
Aime 25(MAA (Mathematical Association of America))
14.0
Math Index(Artificial Analysis)
14.0
Intelligence Index(Artificial Analysis)
7.7
Terminalbench Hard(Stanford × Laude Institute (2026))
4.5
Hle(Center for AI Safety + Scale AI (2025))
3.6

LLM Stats 分類評分

(LLM Stats (zeroeval))
Chat
90
Roleplay
90
Communication
90
Creativity
90
Instruction Following
80
Language
80
Math
80
Reasoning
80
Structured Output
80
General
80
Writing
80
Legal
70
Finance
70
Healthcare
70
Code
70
Physics
50
Biology
50
Chemistry
50

定價

輸入價格$0.475 / 1M tokens
輸出價格$0.495 / 1M tokens
混合價格(3:1)$0.48 / 1M tokens

速度

Tokens/秒0.0
首Token延遲0.00s
首回答延遲0.00s

供應商價格排行

供應商價格排行

3 個供應商

最便宜: DeepInfra最貴: Alibaba (China)
供應商輸入輸出
1DeepInfra最便宜
$0
$0
2Alibaba主要
$0.475
$0.495
3Alibaba (China)
$0.574
$1.721

比較該模型在不同 API 供應商之間的定價。

外部連結

LLM StatsArtificial Analysis