Qwen3.7 Max
AlibabaQwenProprietary
Descripción
Qwen3.7 Max is Alibaba Cloud Qwen Team's proprietary flagship model for agent-driven workflows. It is designed for coding agents, office automation, MCP and multi-agent orchestration, and long-horizon autonomous execution, with a 1 million token context window and up to 65,536 output tokens. Qwen reports strong agentic coding results including 69.7 on Terminal-Bench 2.0-Terminus, 80.4 on SWE-bench Verified, 60.6 on SWE-Pro, and 78.3 on SWE-Multilingual, alongside 92.4 on GPQA Diamond and 97.1 on HMMT 2026 Feb.
Fecha de lanzamiento
2026-05-19
Parámetros
—
Longitud del contexto
1.0M
Modalidades
text
Radar de capacidades
45
general
63
coding
92
reasoning
67
science
70
agents
0
multimodal
Rankings
| Dominio | #Posición | Puntuación | Fuente |
|---|---|---|---|
| Capacidad agéntica | 5 | 66.0 | LS |
| Ranking de codificación | 55 | 84.0 | AA |
| Ranking general | 31 | 85.0 | AA |
| Razonamiento matemático | 30 | 85.0 | LB |
| Razonamiento | 26 | 83.0 | LB |
| Ciencia | 48 | 83.0 | AA |
Puntuaciones de benchmarks (LLM Stats)
(LLM Stats (zeroeval))Agents
QwenSVG
1608.00 / 2000Aut.
Kernel Bench L3
96.0%Aut.
SpreadSheetBench-v1
87.0%Aut.
CoWorkBench
67.2%Aut.
MCP-Mark
60.8%Aut.
QwenWorldBench
57.3%Aut.
Finance Agent v2
48.4%
VITA-Bench
47.9%Aut.
Chat
IFEvalGoogle Research (2023)
94.3%Aut.
Code
QwenWebBench
1568.00 / 2000Aut.
Claw-Eval
65.2%Aut.
ZClawBench
64.3%Aut.
SkillsBench
59.2%Aut.
NL2Repo
47.2%Aut.
General
MAXIFE
89.2%Aut.
Include
86.2%Aut.
NOVA-63
59.0%Aut.
Instruction Following
IFBench
79.1%Aut.
Language
MMLU-Redux
95.0%Aut.
MMMLU
90.3%Aut.
MMLU-Pro
89.6%Aut.
MMLU-ProX
87.0%Aut.
WMT24++
85.8%Aut.
Long Context
MRCR 128K (8-needle)
90.4%Aut.
Math
HMMT Feb 26
97.1%Aut.
IMO-AnswerBench
90.0%Aut.
PolyMATH
86.5%Aut.
LiveBench
74.3%
MathArena Apex
44.5%Aut.
Reasoning
GPQANYU + Cohere + Anthropic (2023)
92.4%Aut.
LiveCodeBench v6
91.6%Aut.
Global PIQA
91.4%Aut.
SWE-Bench Verified
80.4%Aut.
SWE-bench Multilingual
78.3%Aut.
MCP Atlas
76.4%Aut.
SuperGPQA
73.6%Aut.
Terminal-Bench 2.0Stanford × Laude Institute (2026)
69.7%Aut.
SWE-Bench ProPrinceton NLP (2024)
60.6%Aut.
SciCode
53.5%Aut.
Humanity's Last Exam
41.4%Aut.
CritPT
11.4%Aut.
Tool Calling
BFCL-V4
75.0%Aut.
Índices de evaluación AA
(Artificial Analysis)Tau2(Sierra + U Toronto + Vector Institute (2025))94.7
Gpqa(NYU + Cohere + Anthropic (2023))92.3
Ifbench(Google Research (2023))80.5
Lcr(Artificial Analysis)74.7
Terminalbench V2 174.5
Coding Index(Artificial Analysis)66.0
Terminalbench Hard(Stanford × Laude Institute (2026))50.8
Scicode(UIUC + Argonne National Lab (2024))48.8
Intelligence Index(Artificial Analysis)46.7
Hle(Center for AI Safety + Scale AI (2025))40.5
Tau Banking11.8
Puntuaciones por categoría LLM Stats
(LLM Stats (zeroeval))Chat90
Language90
Multimodal90
Spatial Reasoning90
Structured Output90
Instruction Following90
Legal80
Physics80
Productivity80
Frontend Development80
Healthcare80
Math70
Reasoning70
Finance70
General70
Biology70
Chemistry70
Code70
Economics70
Tool Calling70
Agents60
Vision60
Precios
Precio de entrada$2.5 / 1M tokens
Precio de salida$7.5 / 1M tokens
Precio mixto (3:1)$3.75 / 1M tokens
Precio de lectura caché$0.5 / 1M tokens
Precio de escritura caché$3.125 / 1M tokens
Velocidad
Tokens/seg0.0
Retraso del primer token0.00s
Tiempo hasta la respuesta0.00s
Ranking de Precios por Proveedor
Ranking de Precios por Proveedor
24 proveedores
Más barato: NovitaMás caro: Modelis
ProveedorEntradaSalida
1NovitaMás barato
$0
$0
2Together
$0
$0.00001
3Merge Gateway
$0.825
$2.4755
4NovitaAI
$1.25
$3.75
5Kilo Gateway
$1.25
$3.75
6DevPass (LLM Gateway)
$1.25
$3.75
7OrcaRouter
$1.25
$3.75
8Pioneer
$1.25
$3.75
9OpenRouter
$1.475
$4.425
10CrossModel
$1.504
$4.504
11AIHubMix
$1.69
$5.07
12AlibabaPRINCIPAL
$2.5
$7.5
13NanoGPT
$2.5
$7.5
14Abacus
$2.5
$7.5
15OpenCode Go
$2.5
$7.5
16Alibaba (China)
$2.5
$7.5
17ZenMux
$2.5
$7.5
18Alibaba Coding Plan
$2.5
$7.5
19Requesty
$2.5
$7.5
20Alibaba Coding Plan (China)
$2.5
$7.5
21EmpirioLabs AI
$2.5
$7.5
22Charm Hyper
$2.5
$7.5
23Impossibl
$2.5
$7.5
24Modelis
$3
$9
Comparar precios entre diferentes proveedores de API para este modelo.