Qwen 3.6 Plus is Alibaba's latest flagship language model released in April 2026, delivering significant advances in agentic coding, multimodal code generation, and 1M-token context handling. As a native multimodal model, it can generate front-end web pages from screenshots, design drafts, and text prompts, and autonomously breaks down complex programming tasks through iterative write-test-debug cycles. It matches Claude Opus 4.5 on programming and agent benchmarks including SWE-bench and Terminal-Bench 2.0, making it a strong contender for enterprise agentic AI deployment.
VisionReasoning|Proprietary Model
Knowledge Cutoff
Unknown
Input → Output Format
Context Memory
1MIN66KOUT
AI Performance Evaluation
Arena Overall Score
1447
±8As of 2026-04-23
Overall Rank
No.41
5,480 Votes
Arena by Ability
Hard Prompts
1466±10No.39
Expert Knowledge
1483±27No.27
Instruction Following
1434±14No.39
Conversation Memory
1434±19No.56
Creative
1390±20No.76
Coding
1502±15No.30
Math
1484±29No.7
Arena by Occupation
Creative Writing
1420±16No.46
Social Sciences
1447±19No.63
Media
1400±18No.55
Business
1457±18No.27
Healthcare
1472±32No.37
Legal
1445±30No.56
Software
1484±13No.39
Mathematics
1451±31No.33
Source:Arena Intelligence
Overall
AA Intelligence Index
50%↑12%
LiveBench
71%↑10%
Reasoning & Math
GPQA Diamond
88%↑7%
HLE
26%↑9%
LB Reasoning
76%↑16%
LB Math
84%↑10%
LB Data
70%↑20%
Coding
AA Coding Index
43%↑9%
LB Coding
78%↑5%
LB Agentic
55%↑12%
TAU2
98%↑24%
TerminalBench
44%↑13%
SciCode
41%↑0%
Language & Instructions
IFBench
75%↑18%
AA-LCR
70%↑8%
LB Language
75%↑3%
LB IF
58%↑12%
Output Speed
Standard Mode
53tok/s↓29
First Output 106.67s
