Qwen3.7 Plus
AlibabaQwenProprietary
Description
Qwen3.7-Plus is Alibaba Cloud Qwen Team's multimodal agent model that unifies vision and language into a single agent foundation. Built on the Qwen3.7 text backbone, it operates as a multimodal interactive hybrid agent—perceiving real-world scenes, reading screens and operating GUIs, writing code from visual references, navigating mobile apps end-to-end, and answering search-augmented visual questions—while blending GUI and CLI interactions within a single agent loop. It is a versatile coding agent and productivity assistant with full-modality input, generalizing across scaffolds such as Claude Code, OpenClaw, and Qwen Code. Features a 1 million token context window, up to 65,536 output tokens, always-on thinking, and a preserve_thinking mode for agentic tasks. Available via Alibaba Cloud Model Studio (DashScope).
Date de sortie
2026-06-01
Paramètres
—
Longueur du contexte
1.0M
Modalités
image, text, video
Radar de capacités
27
general
54
coding
90
reasoning
64
science
70
agents
90
multimodal
Classements
| Domaine | #Rang | Score | Source |
|---|---|---|---|
| Capacité agentique | 38 | 54.0 | LS |
| Classement codage | 103 | 73.0 | AA |
| Classement général | 71 | 67.0 | AA |
| Classement multimodal | 3 | 74.0 | LS |
| Science | 76 | 72.0 | AA |
Scores de benchmarks (LLM Stats)
(LLM Stats (zeroeval))Agents
SpreadSheetBench-v1
86.3%Aut.
AndroidWorld
81.0%Aut.
CoWorkBench
65.1%Aut.
DeepPlanning
62.3%Aut.
QwenWorldBench
62.1%Aut.
QwenClawBench
61.8%Aut.
MCP-Mark
58.7%Aut.
ClawEval-MM
55.7%Aut.
VITA-Bench
45.6%Aut.
Finance Agent v2
38.2%
Chat
IFEvalGoogle Research (2023)
94.6%Aut.
Code
Claw-Eval
62.7%Aut.
SkillsBench
54.9%Aut.
NL2Repo
41.1%Aut.
General
MAXIFE
88.8%Aut.
Include
83.0%Aut.
NOVA-63
58.8%Aut.
Instruction Following
IFBench
79.1%Aut.
Language
MMLU-Redux
94.5%Aut.
MMMLU
89.0%Aut.
MMLU-Pro
88.5%Aut.
MMLU-ProX
85.4%Aut.
WMT24++
84.6%Aut.
Long Context
MRCR v2
91.7%Aut.
Math
HMMT Feb 26
92.9%Aut.
MathVision
90.3%Aut.
IMO-AnswerBench
86.0%Aut.
PolyMATH
84.0%Aut.
Multimodal
Video-MME
88.0%Aut.
VideoMMMU
85.4%Aut.
OSWorld-Verified
73.3%Aut.
MMSearch-Plus
41.4%Aut.
Reasoning
Global PIQA
90.3%Aut.
GPQANYU + Cohere + Anthropic (2023)
90.3%Aut.
LiveCodeBench v6
89.6%Aut.
CharXiv-R
85.9%Aut.
SWE-Bench Verified
77.7%Aut.
SWE-bench Multilingual
75.8%Aut.
MCP Atlas
73.2%Aut.
SuperGPQA
71.4%Aut.
Terminal-Bench 2.0Stanford × Laude Institute (2026)
70.3%Aut.
SWE-Bench ProPrinceton NLP (2024)
57.6%Aut.
SciCode
51.3%Aut.
Humanity's Last Exam
34.7%Aut.
Apex
22.7%Aut.
FrontierCode 1.1
10.2%
CritPT
6.0%Aut.
Tool Calling
BFCL-V4
72.9%Aut.
Video
MLVU
87.4%Aut.
Vision
OmniDocBench 1.5
91.4%Aut.
RealWorldQA
86.9%Aut.
HiPhO
84.1%Aut.
LingoQA
83.4%Aut.
SimpleVQA
0.82 / 100Aut.
ScreenSpot Pro
79.0%Aut.
MMMU-Pro
79.0%Aut.
TVBench
78.2%Aut.
VLADBench
77.2%Aut.
SURDS
77.2%Aut.
CountQA
77.0%Aut.
LVBench
76.2%Aut.
MedXpertQA-MM
71.0%Aut.
BabyVision
70.4%Aut.
ERQA
69.8%Aut.
OCRBench_V2
67.1%Aut.
WorldVQA
61.1%Aut.
ODinW
51.1%Aut.
BC-VL
51.1%Aut.
MMBC
46.3%Aut.
VisFactor
42.8%Aut.
Indices d'évaluation AA
(Artificial Analysis)Tau2(Sierra + U Toronto + Vector Institute (2025))93.0
Gpqa(NYU + Cohere + Anthropic (2023))90.0
Ifbench(Google Research (2023))78.0
Lcr(Artificial Analysis)73.0
Terminalbench V2 161.0
Coding Index(Artificial Analysis)55.9
Terminalbench Hard(Stanford × Laude Institute (2026))47.0
Scicode(UIUC + Argonne National Lab (2024))46.1
Hle(Center for AI Safety + Scale AI (2025))35.6
Intelligence Index(Artificial Analysis)25.2
Tau Banking17.5
Terminalbench V4 01.0
Scores par catégorie LLM Stats
(LLM Stats (zeroeval))Chat90
Instruction Following90
Language90
Long Context90
Structured Output90
Video90
Legal80
Multimodal80
Physics80
Productivity80
Spatial Reasoning80
Frontend Development80
Grounding80
Healthcare80
Image To Text70
Math70
Reasoning70
Finance70
General70
Biology70
Chemistry70
Economics70
Tool Calling70
Vision70
Agents60
Code60
Tarification
Prix d'entrée$0.4 / 1M tokens
Prix de sortie$1.6 / 1M tokens
Prix mixte (3:1)$0.7 / 1M tokens
Prix de lecture cache$0.08 / 1M tokens
Prix d'écriture cache$0.5 / 1M tokens
Vitesse
Tokens/sec56.7
Délai du premier token1.67s
Temps de réponse36.92s
Classement des Prix par Fournisseur
Classement des Prix par Fournisseur
22 fournisseurs
Moins cher: 302.AIPlus cher: Charm Hyper
FournisseurEntréeSortie
1302.AIMoins cher
$0.228
$0.92
2Alibaba (China)
$0.276
$1.101
3AIHubMix
$0.282
$1.128
4OpenRouter
$0.32
$1.28
5Kilo Gateway
$0.32
$1.28
6Requesty
$0.32
$1.28
7CrossModel
$0.32
$1.25
8Standard Compute
$0.32
$1.28
9OrcaRouter
$0.35
$1.42
10Kenari
$0.38286
$1.48571
11AlibabaPRINCIPAL
$0.4
$1.6
12NanoGPT
$0.4
$1.6
13OpenCode Go
$0.4
$1.6
14ZenMux
$0.4
$1.6
15DevPass (LLM Gateway)
$0.4
$1.6
16Merge Gateway
$0.4
$1.6
17LLMTR
$0.4
$1.6
18EmpirioLabs AI
$0.4
$1.6
19Ofox
$0.4
$1.6
20Impossibl
$0.4
$1.6
21Modelis
$0.768
$3.072
22Charm Hyper
$1.2
$4.8
Comparer les prix entre différents fournisseurs API pour ce modèle.