跳转到主要内容

Qwen3.5 0.8B (Non-reasoning)

AlibabaQwen开源权重Apache 2.0 · 商用许可

描述

Qwen3.5-0.8B is a 0.8 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 features both thinking and non-thinking modes.

发布日期
2026-03-02
参数规模
800M
上下文长度
支持模态

能力雷达图

3
general
1
coding
24
reasoning
13
science
20
agents
10
multimodal

排行榜排名

领域#排名分数来源
智能体能力模型榜144
27.0
LS
代码能力榜539
4.0
AA
通用能力榜505
21.0
AA
科学能力567
9.0
AA

基准测试分数 (LLM Stats)

(LLM Stats (zeroeval))

Agents

BFCL-V425.3%自报
t2-bench11.6%自报

Biology

GPQANYU + Cohere + Anthropic (2023)11.9%自报

Chemistry

SuperGPQA21.3%自报

Communication

Multi-Challenge18.9%自报

Finance

MMLU-Pro42.3%自报
MMLU-ProX34.6%自报

General

MMLU-Redux59.5%自报
Global PIQA59.4%自报
C-Eval50.5%自报
MMMLU44.3%自报
IFEvalGoogle Research (2023)44.0%自报
NOVA-6342.4%自报
Include40.6%自报
MAXIFE39.2%自报
LongBench v226.1%自报
IFBench21.0%自报

Language

WMT24++27.2%自报

Long Context

AA-LCR4.7%自报

Math

PolyMATH8.2%自报

AA 评测指数

(Artificial Analysis)
Intelligence Index(Artificial Analysis)
2.9
Coding Index(Artificial Analysis)
1.2
Tau2(Sierra + U Toronto + Vector Institute (2025))
0.7
Gpqa(NYU + Cohere + Anthropic (2023))
0.2
Ifbench(Google Research (2023))
0.2
Lcr(Artificial Analysis)
0.1
Hle(Center for AI Safety + Scale AI (2025))
0.1
Scicode(UIUC + Argonne National Lab (2024))
0.0
Terminalbench V2 1
0.0
Terminalbench Hard(Stanford × Laude Institute (2026))
0.0

LLM Stats 分类评分

(LLM Stats (zeroeval))
Math
40
Structured Output
40
Language
40
Legal
30
Physics
30
Reasoning
30
Instruction Following
30
Finance
30
General
30
Healthcare
30
Long Context
20
Agents
20
Chemistry
20
Communication
20
Economics
20
Tool Calling
20
Multimodal
10
Spatial Reasoning
10
Biology
10
Vision
10

定价

输入价格免费
输出价格免费
混合价格(3:1)免费

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

暂无提供商数据

外部链接