Skip to main content
AITier

Qwen3.5 2B (Reasoning)

AlibabaQwenOpen WeightApache 2.0 · Commercial OK

Description

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.

Release Date
2026-03-02
Parameters
2.0B
Context Length
—
Modalities
—

Capability Radar

13
general
3
coding
46
reasoning
22
scienceest.
50
agents
30
multimodal

Science uses a reasoning proxy when dedicated science benchmarks are unavailable.

Rankings

Domain#RankScoreSource
Agents & Tools80
47.0
LS
Code Ranking409
12.0
AA
General Ranking315
35.0
AA
Science441
16.0
AA

Benchmark Scores (LLM Stats)

Agents

t2-bench48.8%SR
BFCL-V443.6%SR

Biology

GPQA51.6%SR

Chemistry

SuperGPQA37.5%SR

Communication

Multi-Challenge33.7%SR

Finance

MMLU-Pro66.5%SR
MMLU-ProX52.3%SR

General

MMLU-Redux79.6%SR
IFEval78.6%SR
C-Eval73.2%SR
Global PIQA69.3%SR
MMMLU63.1%SR
MAXIFE60.6%SR
Include55.4%SR
NOVA-6346.4%SR
IFBench41.3%SR
LongBench v238.7%SR

Language

WMT24++45.8%SR

Long Context

AA-LCR25.6%SR

Math

PolyMATH26.1%SR

AA Evaluation Indices

Intelligence Index
16.3
Coding Index
3.5
Tau2
0.7
Gpqa
0.5
Ifbench
0.3
Lcr
0.2
Terminalbench Hard
0.0
Scicode
0.0
Hle
0.0

LLM Stats Category Scores

Structured Output
60
General
60
Instruction Following
60
Language
60
Tool Calling
50
Agents
50
Biology
50
Finance
50
Healthcare
50
Legal
50
Math
50
Physics
50
Reasoning
50
Chemistry
40
Economics
40
Spatial Reasoning
30
Vision
30
Communication
30
Long Context
30
Multimodal
30

Pricing

Input Price$0.02 / 1M tokens
Output Price$0.1 / 1M tokens
Blended Price (3:1)$0.04 / 1M tokens

Speed

Tokens/sec0.0 tokens/s
Time to First Token0.00s
Time to Answer0.00s

Available Providers

(LS internal units)

No provider data available

External Sources

LLM Stats