Qwen3.5 0.8B (Non-reasoning)
AlibabaQwenOpen WeightApache 2.0 · Commercial OK
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
41K
支持模态
text
能力雷达图
9
general
1
coding
24
reasoning
13
science估算
20
agents
0
multimodal
Science 在缺少专门科学评测时使用推理能力代理估算。
排行榜排名
基准测试分数 (LLM Stats)
Agents
BFCL-V4
25.3%自报
t2-bench
11.6%自报
Biology
GPQA
11.9%自报
Chemistry
SuperGPQA
21.3%自报
Communication
Multi-Challenge
18.9%自报
Finance
MMLU-Pro
42.3%自报
MMLU-ProX
34.6%自报
General
MMLU-Redux
59.5%自报
Global PIQA
59.4%自报
C-Eval
50.5%自报
MMMLU
44.3%自报
IFEval
44.0%自报
NOVA-63
42.4%自报
Include
40.6%自报
MAXIFE
39.2%自报
LongBench v2
26.1%自报
IFBench
21.0%自报
Language
WMT24++
27.2%自报
Long Context
AA-LCR
4.7%自报
Math
PolyMATH
8.2%自报
AA 评测指数
Intelligence Index9.9
Coding Index1.0
Tau20.7
Gpqa0.2
Ifbench0.2
Lcr0.1
Hle0.0
Scicode0.0
Terminalbench Hard0.0
LLM Stats 分类评分
Structured Output40
General40
Language40
Math40
Finance30
Healthcare30
Instruction Following30
Legal30
Physics30
Reasoning30
Tool Calling20
Agents20
Chemistry20
Communication20
Economics20
Long Context20
Spatial Reasoning10
Vision10
Biology10
Multimodal10
定价
输入价格$0.01 / 1M tokens
输出价格$0.05 / 1M tokens
混合价格(3:1)$0.02 / 1M tokens
速度
Tokens/秒367.0 tokens/s
首Token延迟0.48s
首回答延迟0.48s
可用提供商
(LS 内部计价单位)暂无提供商数据