Qwen3.5 2B (Reasoning)
AlibabaQwenOpen WeightApache 2.0 · Commercial OK
描述
Qwen3.5-2B is a 2 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
2.0B
上下文长度
—
支持模态
—
能力雷达图
13
general
3
coding
46
reasoning
22
science估算
50
agents
30
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
t2-bench
48.8%自报
BFCL-V4
43.6%自报
Biology
GPQA
51.6%自报
Chemistry
SuperGPQA
37.5%自报
Communication
Multi-Challenge
33.7%自报
Finance
MMLU-Pro
66.5%自报
MMLU-ProX
52.3%自报
General
MMLU-Redux
79.6%自报
IFEval
78.6%自报
C-Eval
73.2%自报
Global PIQA
69.3%自报
MMMLU
63.1%自报
MAXIFE
60.6%自报
Include
55.4%自报
NOVA-63
46.4%自报
IFBench
41.3%自报
LongBench v2
38.7%自报
Language
WMT24++
45.8%自报
Long Context
AA-LCR
25.6%自报
Math
PolyMATH
26.1%自报
AA 评测指数
Intelligence Index16.3
Coding Index3.5
Tau20.7
Gpqa0.5
Ifbench0.3
Lcr0.2
Terminalbench Hard0.0
Scicode0.0
Hle0.0
LLM Stats 分类评分
Structured Output60
General60
Instruction Following60
Language60
Tool Calling50
Agents50
Biology50
Finance50
Healthcare50
Legal50
Math50
Physics50
Reasoning50
Chemistry40
Economics40
Spatial Reasoning30
Vision30
Communication30
Long Context30
Multimodal30
定价
输入价格$0.02 / 1M tokens
输出价格$0.1 / 1M tokens
混合价格(3:1)$0.04 / 1M tokens
速度
Tokens/秒0.0 tokens/s
首Token延迟0.00s
首回答延迟0.00s
可用提供商
(LS 内部计价单位)暂无提供商数据