Qwen3.6 Plus

Qwen3.6 Plus is Alibaba's next-generation flagship model featuring a 1 million token native context window, up to 65,536 output tokens, and always-on chain-of-thought reasoning. It uses a next-generation hybrid architecture optimized for efficiency and scalability.

CountBench

97.6%

i
V*

96.9%

i
HMMT 2025

96.7%

i
AIME 2026

95.3%

i
HMMT25

94.6%

i
MMLU-Redux

94.5%

i
AI2D

94.4%

i
IFEval

94.3%

i
RefCOCO-avg

93.5%

i
C-Eval

93.3%

i
OmniDocBench 1.5

91.2%

i
GPQA

90.4%

i
Global PIQA

89.8%

i
MMMLU

89.5%

i
We-Math

89.0%

i
MMLU-Pro

88.5%

i
MAXIFE

88.2%

i
DynaMath

88.0%

i
MathVision

88.0%

i
HMMT Feb 26

87.8%

i
LiveCodeBench v6

87.1%

i
MLVU

86.7%

i
MMMU

86.0%

i
RealWorldQA

85.4%

i
Include

85.1%

i
MMLU-ProX

84.7%

i
WMT24++

84.3%

i
Video-MME

84.2%

i
VideoMMMU

84.0%

i
IMO-AnswerBench

83.8%

i
CC-OCR

83.4%

i
MMStar

83.3%

i
CharXiv-R

81.5%

i
MMMU-Pro

78.8%

i
SWE-Bench Verified

78.8%

i
PolyMATH

77.4%

i
WideSearch

74.3%

i
IFBench

74.2%

i
MCP Atlas

74.1%

i
SWE-bench Multilingual

73.8%

i
SuperGPQA

71.6%

i
TAU3-Bench

70.7%

i
AA-LCR

68.3%

i
ScreenSpot Pro

68.2%

i
SimpleVQA

67.3%

i
ERQA

65.7%

i
OSWorld-Verified

62.5%

i
LongBench v2

62.0%

i
MMLongBench-Doc

62.0%

i
Terminal-Bench 2.0

61.6%

i
TIR-Bench

61.6%

i
Claw-Eval

58.7%

i
NOVA-63

57.9%

i
SWE-Bench Pro

56.6%

i
ODinW

51.8%

i
MCP-Mark

48.2%

i
SkillsBench

45.7%

i
VITA-Bench

44.3%

i
DeepPlanning

41.5%

i
Toolathlon

39.8%

i
NL2Repo

37.9%

i
Humanity's Last Exam

28.8%

i

Pricing, uptime, and speed via OpenRouter — updated Jul 17, 2026, 04:19 AM.

Provider	Status	Input	Output	Limits	Uptime	Speed	Notes
Alibaba	available	$0.33/Mtok	$1.95/Mtok	1.0M tokens context 66K tokens max output	100.0% 5m 100.0%	980 ms p50 TTFT 42 tok/s p50