跳转到主要内容
AITier

Qwen3.8 Flash

Alibaba Cloud / Qwen TeamQwenProprietary

描述

Qwen3.8 Flash is the production QwenCloud / OpenRouter API model (id qwen3.8-flash), not the open-weight Qwen3.8-Flash-Next checkpoint. Hugging Face states Flash is the official managed version based on Flash-Next with production features such as 1M context by default and official built-in tools. Architecture inherits Flash-Next (125B parameters with 6B activated, plus 51B n-gram embedding and 4B MTP). Multimodal text, image, and video understanding; thinking on by default; function calling, built-in tools, and structured output. QwenCloud lists 1M context, ~128k max output, and ~256k thinking budget. Open weights remain under the separate catalog card qwen3.8-flash-next.

发布日期
2026-08-26
参数规模
125.0B
上下文长度
1.0M
支持模态
image, text, video

能力雷达图

70
general
60
coding
70
reasoning
77
science估算
60
agents
70
multimodal

缺少专门科学评测时，Science 由 LLM Stats 科学得分或推理能力估算。

排行榜排名

领域#排名分数来源
智能体能力模型榜4
67.0
LS
多模态榜33
60.0
LS

基准测试分数 (LLM Stats)

(LLM Stats (zeroeval))

Agents

AndroidWorld84.5%自报
CoWorkBench73.9%自报
Toolathlon73.5%自报
ClawEval-MM60.4%自报
Job Bench55.7%自报
Agents' Last Exam51.2%自报
RecreationBench49.9%自报

Code

Vision2Web64.0%自报
DeepSWE 1.158.7%自报
NL2Repo48.1%自报

Instruction Following

IFBench81.3%自报

Math

MathVision95.7%自报

Multimodal

OSWorld 2.019.4%自报

Reasoning

LiveCodeBench v691.9%自报
GPQANYU + Cohere + Anthropic (2023)91.7%自报
CharXiv-R90.6%自报
SWE-bench Multilingual81.0%自报
SWE-Bench ProPrinceton NLP (2024)62.5%自报
Humanity's Last Exam35.9%自报

Vision

RealWorldQA88.5%自报
LVBench76.6%自报
ERQA72.3%自报

AA 评测指数

(Artificial Analysis)

暂无 AA 评测数据

LLM Stats 分类评分

(LLM Stats (zeroeval))
Physics
90
Biology
90
Chemistry
90
Long Context
80
Spatial Reasoning
80
Instruction Following
80
Math
70
Multimodal
70
Reasoning
70
General
70
Vision
70
Productivity
60
Agents
60
Code
60
Tool Calling
60

定价

输入价格$0 / 1M tokens
输出价格$0 / 1M tokens
混合价格(3:1)$0 / 1M tokens
缓存读取价格$0.016 / 1M tokens
缓存写入价格$0.2 / 1M tokens

速度

暂无速度数据

供应商价格排行

供应商价格排行

12 个供应商

最便宜: Alibaba Cloud / Qwen Team最贵: Eden AI
供应商输入输出
1Alibaba Cloud / Qwen Team主要
$0
$0
2Novita
$0
$0
3Alibaba (China)
$0.11875
$0.40073
4Vancine
$0.12
$0.38
5CrossModel
$0.13
$0.43
6Alibaba
$0.15
$0.47
7OpenRouter
$0.15
$0.47
8OpenCode Go
$0.15
$0.47
9Kilo Gateway
$0.15
$0.47
10DevPass (LLM Gateway)
$0.15
$0.47
11Charm Hyper
$0.15
$0.47
12Eden AI
$0.16
$0.47

比较该模型在不同 API 供应商之间的定价。

外部链接

LLM StatsArtificial Analysis