Qwen3 8B (Non-reasoning)

AlibabaQwen

Fecha de lanzamiento

2025-04-28

Parámetros

—

Longitud del contexto

262K

Modalidades

audio, image, text, video

Radar de capacidades

general

coding

reasoning

scienceest.

agents

multimodal

Science usa un proxy de razonamiento cuando los benchmarks científicos dedicados no están disponibles.

Rankings

Dominio	#Posición	Puntuación	Fuente
Ranking de codificación	457	9.0	AA
Ranking general	404	26.0	AA
Razonamiento matemático	244	36.0	AA
Ciencia	418	24.0	AA

Puntuaciones de benchmarks (LLM Stats)

3d

SUNRGBD

0.36 / 100Aut.

Hypersim

0.13 / 100Aut.

Agents

GDPval-AA

985.00 / 3000Aut.

t2-bench

79.5%Aut.

BFCL-V4

72.2%Aut.

AndroidWorld_SR

66.4%Aut.

BrowseComp

63.8%Aut.

FullStackBench en

62.6%Aut.

WideSearch

60.5%Aut.

FullStackBench zh

58.7%Aut.

OSWorld-Verified

58.0%Aut.

TIR-Bench

53.2%Aut.

Terminal-Bench 2.0

49.4%Aut.

VITA-Bench

33.6%Aut.

DeepPlanning

24.1%Aut.

Biology

GPQA

86.6%Aut.

Chemistry

SuperGPQA

67.1%Aut.

Code

SWE-Bench Verified

72.0%Aut.

Communication

Multi-Challenge

61.5%Aut.

Embodied

EmbSpatialBench

0.84 / 100Aut.

Finance

MMLU-Pro

86.7%Aut.

MMLU-ProX

82.2%Aut.

General

MMLU-Redux

94.0%Aut.

IFEval

93.4%Aut.

C-Eval

91.9%Aut.

Global PIQA

88.4%Aut.

MAXIFE

87.9%Aut.

MMMLU

86.7%Aut.

MMMU

83.9%Aut.

MMStar

82.9%Aut.

Include

82.8%Aut.

LiveCodeBench v6

78.9%Aut.

MMMU-Pro

76.9%Aut.

IFBench

76.1%Aut.

SimpleVQA

0.62 / 100Aut.

LongBench v2

60.2%Aut.

NOVA-63

58.6%Aut.

Grounding

RefCOCO-avg

0.91 / 100Aut.

ScreenSpot Pro

70.4%Aut.

RefSpatialBench

0.69 / 100Aut.

Healthcare

VideoMMMU

82.0%Aut.

SlakeVQA

81.6%Aut.

MedXpertQA

67.3%Aut.

PMC-VQA

63.3%Aut.

Image To Text

OCRBench

92.1%Aut.

Language

LingoQA

80.8%Aut.

WMT24++

78.3%Aut.

Long Context

MLVU

87.3%Aut.

LVBench

74.4%Aut.

AA-LCR

66.9%Aut.

MMLongBench-Doc

0.59 / 100Aut.

Math

HMMT 2025

91.4%Aut.

HMMT25

90.3%Aut.

MathVista-Mini

87.4%Aut.

MathVision

86.2%Aut.

DynaMath

85.9%Aut.

CodeForces

0.85 / 3000Aut.

PolyMATH

68.9%Aut.

Humanity's Last Exam

47.5%Aut.

Multimodal

VLMsAreBlind

96.7%Aut.

AI2D

93.3%Aut.

93.2%Aut.

MMBench-V1.1

92.8%Aut.

OmniDocBench 1.5

89.8%Aut.

VideoMME w sub.

87.3%Aut.

VideoMME w/o sub.

83.9%Aut.

CC-OCR

81.8%Aut.

CharXiv-R

77.2%Aut.

MVBench

76.6%Aut.

MMVU

74.7%Aut.

BabyVision

40.2%Aut.

ZEROBench-Sub

0.36 / 100Aut.

Nuscene

15.4%Aut.

ZEROBench

0.09 / 100Aut.

Reasoning

CountBench

0.97 / 100Aut.

BrowseComp-zh

69.9%Aut.

Hallusion Bench

67.6%Aut.

ERQA

62.0%Aut.

Seal-0

44.1%Aut.

OJBench

39.5%Aut.

Spatial Reasoning

RealWorldQA

85.1%Aut.

Vision

ODinW

44.5%Aut.

Índices de evaluación AA

Math Index

24.3

Intelligence Index

5.1

Math 500

0.8

Mmlu Pro

0.6

Gpqa

0.5

Ifbench

0.3

Tau2

0.2

Aime

0.2

Aime 25

0.2

Livecodebench

0.2

Scicode

0.2

Hle

0.0

Terminalbench Hard

0.0

Lcr

0.0

Puntuaciones por categoría LLM Stats

Legal

100

Finance

100

Agents

General

Reasoning

Biology

Image To Text

Instruction Following

Language

Math

Physics

Structured Output

Embodied

Grounding

Healthcare

Chemistry

Text-to-image

Video

Long Context

Multimodal

Spatial Reasoning

Frontend Development

Economics

Vision

Code

Communication

Tool Calling

Spatial

Precios

Precio de entrada$0.18 / 1M tokens

Precio de salida$0.2 / 1M tokens

Precio mixto (3:1)$0.185 / 1M tokens

Velocidad

Tokens/seg63.5

Retraso del primer token1.44s

Tiempo hasta la respuesta1.44s

Ranking de Precios por Proveedor

1 proveedores

ProveedorEntradaSalida

1AlibabaPRINCIPAL

$0.18

$0.2

Comparar precios entre diferentes proveedores de API para este modelo.

Fuentes externas

Artificial Analysis