跳转到主要内容
AITier

Qwen3.8 Max

AlibabaQwen开源权重Qwen3.8-Max License · 商用许可

描述

Qwen3.8 Max, released as the Qwen3.8-2.4T-A95B open-weight checkpoint, is Qwen's flagship mixture-of-experts model for coding, research, professional workflows, and long-horizon agents. It has 2.4 trillion total parameters with 95 billion active parameters, a native 262,144-token context window extensible to about 1 million tokens, and always-on thinking. QwenCloud serves the same model in the Qwen3.8 Max API family with text and image input.

发布日期
2026-08-03
参数规模
2.4T
上下文长度
1.0M
支持模态
image, pdf, text, video

能力雷达图

55
general
69
coding
93
reasoning
69
science
60
agents
90
multimodal

排行榜排名

领域#排名分数来源
智能体能力模型榜4
79.0
LS
代码能力榜26
91.0
AA
通用能力榜8
92.0
AA
数学推理17
91.0
LB
多模态榜10
65.0
LS
推理能力13
88.0
LB
科学能力26
87.0
AA

基准测试分数 (LLM Stats)

(LLM Stats (zeroeval))

Agents

QwenReactBench1724.00 / 2000自报
QwenSVG1713.00 / 2000自报
PaperBench93.0%自报
Terminal-Bench 2.186.6%自报
OSWorld-Verified86.1%自报
AndroidWorld85.3%自报
WideSearch81.9%自报
QwenSWEBench80.7%自报
MobileWorld77.8%自报
AndroidBench75.1%自报
CoWorkBench74.8%自报
FrontierSWE73.5%自报
Toolathlon72.5%自报
SkillsBench70.2%自报
Workspace Bench67.7%自报
SWE-Bench ProPrinceton NLP (2024)67.7%自报
QwenQoderBench58.4%自报
DeepSWE 1.156.6%自报
NL2Repo55.9%自报
Job Bench53.4%自报
OneMillion Bench52.5%自报
Agents' Last Exam52.4%自报
MLS-Bench Lite41.0%自报
AutomationBench27.3%自报

Biology

GPQANYU + Cohere + Anthropic (2023)92.6%自报

Code

Vision2Web69.0%自报

Finance

PRBench-Finance58.3%自报

General

MRCR v2 (8-needle)92.9%自报
IFBench82.8%自报
MMMU-Pro82.3%自报
LongBench v266.3%自报

Grounding

ScreenSpot Pro84.5%自报

Healthcare

HealthBench60.2%自报

Knowledge

PLawBench73.2%自报
PRBench-Legal57.6%自报

Long Context

LVBench81.8%自报

Math

Humanity's Last Exam (with tools, text-only)56.2%自报
Humanity's Last Exam43.6%自报

Multimodal

VideoMME w sub.90.4%自报
PerceptionBench63.5%自报

Reasoning

ERQA77.8%自报

Spatial Reasoning

RealWorldQA88.0%自报

AA 评测指数

(Artificial Analysis)
Coding Index(Artificial Analysis)
71.8
Intelligence Index(Artificial Analysis)
58.1
Gpqa(NYU + Cohere + Anthropic (2023))
0.9
Terminalbench V2 1
0.8
Lcr(Artificial Analysis)
0.7
Scicode(UIUC + Argonne National Lab (2024))
0.5
Tau Banking
0.5
Hle(Center for AI Safety + Scale AI (2025))
0.4

LLM Stats 分类评分

(LLM Stats (zeroeval))
Physics
90
Biology
90
Chemistry
90
Video
90
Multimodal
80
Search
80
Spatial Reasoning
80
Instruction Following
80
Grounding
80
Vision
80
Long Context
70
Reasoning
70
Structured Output
70
General
70
Agents
70
Code
70
Productivity
60
Healthcare
60
Tool Calling
60
Math
50

定价

输入价格$2 / 1M tokens
输出价格$6 / 1M tokens
混合价格(3:1)$3 / 1M tokens
缓存读取价格$0.25 / 1M tokens
缓存写入价格$2.5 / 1M tokens

速度

Tokens/秒45.2
首Token延迟1.95s
首回答延迟46.22s

供应商价格排行

供应商价格排行

13 个供应商

最便宜: AIHubMix最贵: Charm Hyper
供应商输入输出
1AIHubMix最便宜
$1.69
$5.07
2Alibaba (China)
$1.77744
$5.33231
3LLM Gateway
$1.815
$5.4461
4CrossModel
$1.88
$5.63
5Alibaba主要
$2
$6
6NanoGPT
$2
$6
7OpenRouter
$2
$6
8OpenCode Go
$2
$6
9Kilo Gateway
$2
$6
10DigitalOcean
$2
$6
11Merge Gateway
$2
$6
12EmpirioLabs AI
$2
$6
13Charm Hyper
$2
$6

比较该模型在不同 API 供应商之间的定价。

外部链接

LLM StatsArtificial Analysis