跳转到主要内容

Qwen3.5 9B (Non-reasoning)

AlibabaQwen开源权重Apache 2.0 · 商用许可

描述

Qwen3.5-9B is a 9 billion parameter vision-language model using Gated DeltaNet hybrid architecture with a 3:1 ratio of linear attention to full softmax attention. It supports 262K native context length and delivers strong performance across knowledge, reasoning, coding, and multilingual tasks.

发布日期
2026-03-02
参数规模
9.0B
上下文长度
262K
支持模态
image, text, video

能力雷达图

12
general
24
coding
79
reasoning
57
science
70
agents
60
multimodal

排行榜排名

领域#排名分数来源
智能体能力模型榜99
35.0
LS
代码能力榜381
35.0
AA
通用能力榜337
40.0
AA
科学能力290
47.0
AA

基准测试分数 (LLM Stats)

(LLM Stats (zeroeval))

Agents

t2-bench79.1%自报
VITA-Bench29.8%自报
DeepPlanning18.0%自报

Chat

IFEvalGoogle Research (2023)91.5%自报
Multi-Challenge54.5%自报

General

C-Eval88.2%自报
MAXIFE83.4%自报
Include75.6%自报
NOVA-6355.9%自报

Instruction Following

IFBench64.5%自报

Language

MMLU-Redux91.1%自报
MMLU-Pro82.5%自报
MMMLU81.2%自报
MMLU-ProX76.3%自报
WMT24++72.6%自报

Long Context

LongBench v255.2%自报

Math

HMMT 202583.2%自报
HMMT2582.9%自报
PolyMATH57.3%自报

Reasoning

Global PIQA83.2%自报
GPQANYU + Cohere + Anthropic (2023)81.7%自报
LiveCodeBench v665.6%自报
AA-LCR63.0%自报
SuperGPQA58.2%自报

Tool Calling

BFCL-V466.1%自报

AA 评测指数

(Artificial Analysis)
Tau2(Sierra + U Toronto + Vector Institute (2025))
85.1
Gpqa(NYU + Cohere + Anthropic (2023))
78.6
Lcr(Artificial Analysis)
46.0
Ifbench(Google Research (2023))
37.8
Coding Index(Artificial Analysis)
23.5
Terminalbench V2 1
21.3
Terminalbench Hard(Stanford × Laude Institute (2026))
18.2
Intelligence Index(Artificial Analysis)
13.3
Hle(Center for AI Safety + Scale AI (2025))
9.4

LLM Stats 分类评分

(LLM Stats (zeroeval))
Instruction Following
80
Language
80
Math
80
Biology
80
Chat
70
Legal
70
Physics
70
Reasoning
70
Structured Output
70
Finance
70
General
70
Healthcare
70
Chemistry
70
Tool Calling
70
Long Context
60
Multimodal
60
Spatial Reasoning
60
Economics
60
Vision
60
Agents
50
Communication
50

定价

输入价格$0.17 / 1M tokens
输出价格$0.25 / 1M tokens
混合价格(3:1)$0.19 / 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

2 个供应商

最便宜: DeepInfra最贵: Alibaba
供应商输入输出
1DeepInfra最便宜
$0
$0
2Alibaba主要
$0.17
$0.25

比较该模型在不同 API 供应商之间的定价。

外部链接