Qwen3.5 35B A3B (Non-reasoning)
AlibabaQwenOpen WeightApache 2.0 · Uso Comercial
Descripción
Qwen3.5-35B-A3B is a multimodal Mixture-of-Experts model with 35 billion total parameters and 3 billion activated parameters. It combines strong reasoning, coding, agentic, and visual understanding performance with production-friendly efficiency and a native 262K context window.
Fecha de lanzamiento
2026-02-24
Parámetros
35.0B
Longitud del contexto
262K
Modalidades
image, text, video
Radar de capacidades
70
general
60
coding
80
reasoning
68
scienceest.
60
agents
80
multimodal
Science se estima a partir de las puntuaciones científicas de LLM Stats o del razonamiento cuando no hay benchmarks científicos dedicados.
Rankings
| Dominio | #Posición | Puntuación | Fuente |
|---|---|---|---|
| Capacidad agéntica | 83 | 46.0 | LS |
| Ranking multimodal | 96 | 46.0 | LS |
Puntuaciones de benchmarks (LLM Stats)
(LLM Stats (zeroeval))Agents
t2-bench
81.2%Aut.
VITA-Bench
31.9%Aut.
DeepPlanning
22.8%Aut.
Chat
IFEvalGoogle Research (2023)
91.9%Aut.
Multi-Challenge
60.0%Aut.
Code
FullStackBench en
58.1%Aut.
FullStackBench zh
55.0%Aut.
General
C-Eval
90.2%Aut.
MAXIFE
86.6%Aut.
Include
79.7%Aut.
AndroidWorld_SR
71.1%Aut.
NOVA-63
57.1%Aut.
Healthcare
PMC-VQA
62.0%Aut.
MedXpertQA
61.4%Aut.
Instruction Following
IFBench
70.2%Aut.
Language
MMLU-Redux
93.3%Aut.
MMLU-Pro
85.3%Aut.
MMMLU
85.2%Aut.
MMLU-ProX
81.0%Aut.
WMT24++
76.3%Aut.
Long Context
LongBench v2
59.0%Aut.
Math
HMMT25
89.2%Aut.
HMMT 2025
89.0%Aut.
MathVista-Mini
86.2%Aut.
DynaMath
85.0%Aut.
MathVision
83.9%Aut.
CodeForces
0.82 / 3000Aut.
PolyMATH
64.4%Aut.
Multimodal
VideoMME w/o sub.
82.5%Aut.
MMMU
81.4%Aut.
VideoMMMU
80.4%Aut.
TIR-Bench
55.5%Aut.
OSWorld-Verified
54.5%Aut.
Reasoning
Global PIQA
86.6%Aut.
GPQANYU + Cohere + Anthropic (2023)
84.2%Aut.
CharXiv-R
77.5%Aut.
LiveCodeBench v6
74.6%Aut.
BrowseComp-zh
69.5%Aut.
SWE-Bench Verified
69.2%Aut.
SuperGPQA
63.4%Aut.
BrowseCompOpenAI (2025)
61.0%Aut.
AA-LCR
58.5%Aut.
Humanity's Last Exam
47.4%Aut.
Seal-0
41.4%Aut.
Terminal-Bench 2.0Stanford × Laude Institute (2026)
40.5%Aut.
OJBench
36.0%Aut.
Search
WideSearch
57.1%Aut.
Tool Calling
BFCL-V4
67.3%Aut.
Video
MLVU
85.6%Aut.
Vision
CountBench
0.98 / 100Aut.
VLMsAreBlind
97.0%Aut.
V*
92.7%Aut.
AI2D
92.6%Aut.
MMBench-V1.1
91.5%Aut.
OCRBench
91.0%Aut.
OmniDocBench 1.5
89.3%Aut.
RefCOCO-avg
0.89 / 100Aut.
VideoMME w sub.
86.6%Aut.
RealWorldQA
84.1%Aut.
EmbSpatialBench
0.83 / 100Aut.
MMStar
81.9%Aut.
CC-OCR
80.7%Aut.
LingoQA
79.2%Aut.
SlakeVQA
78.7%Aut.
MMMU-Pro
75.1%Aut.
MVBench
74.8%Aut.
MMVU
72.3%Aut.
LVBench
71.4%Aut.
ScreenSpot Pro
68.6%Aut.
Hallusion Bench
67.9%Aut.
ERQA
64.8%Aut.
RefSpatialBench
0.64 / 100Aut.
MMLongBench-Doc
0.59 / 100Aut.
SimpleVQA
0.58 / 100Aut.
ODinW
42.6%Aut.
BabyVision
38.4%Aut.
ZEROBench-Sub
0.34 / 100Aut.
SUNRGBD
0.33 / 100Aut.
Nuscene
14.6%Aut.
Hypersim
0.13 / 100Aut.
ZEROBench
0.08 / 100Aut.
Índices de evaluación AA
(Artificial Analysis)No hay datos de evaluación AA disponibles
Puntuaciones por categoría LLM Stats
(LLM Stats (zeroeval))Chat80
Image To Text80
Instruction Following80
Language80
Legal80
Math80
Physics80
Structured Output80
Embodied80
Finance80
Biology80
Text-to-image80
Video80
Long Context70
Multimodal70
Reasoning70
Spatial Reasoning70
Frontend Development70
General70
Grounding70
Healthcare70
Chemistry70
Vision70
Search60
Code60
Communication60
Economics60
Tool Calling60
Agents50
3d20
Spatial10
Precios
Precio de entrada$0.25 / 1M tokens
Precio de salida$2 / 1M tokens
Precio mixto (3:1)$0.355 / 1M tokens
Velocidad
Tokens/seg175.4
Retraso del primer token1.99s
Tiempo hasta la respuesta1.99s
Ranking de Precios por Proveedor
Ranking de Precios por Proveedor
6 proveedores
Más barato: AIHubMixMás caro: DevPass (LLM Gateway)
ProveedorEntradaSalida
1AIHubMixMás barato
$0.0564
$0.4512
2302.AI
$0.06
$0.46
3DeepInfra
$0.14
$1
4Requesty
$0.14
$1
5AlibabaPRINCIPAL
$0.25
$2
6DevPass (LLM Gateway)
$0.25
$2
Comparar precios entre diferentes proveedores de API para este modelo.