Saltar al contenido principal

Qwen3 Next 80B A3B (Reasoning)

AlibabaQwenOpen WeightApache 2.0 · Uso Comercial

Descripción

Qwen3-Next-80B-A3B-Thinking is the thinking variant of the Qwen3-Next series, featuring the same groundbreaking architecture as the instruct model. Leveraging GSPO, it addresses stability and efficiency challenges of hybrid attention + high-sparsity MoE in RL training. It uses Hybrid Attention combining Gated DeltaNet and Gated Attention for efficient ultra-long context modeling, High-Sparsity MoE with 512 experts (10 activated + 1 shared), and Multi-Token Prediction. With 80B total parameters and only 3B activated, it demonstrates outstanding performance on complex reasoning tasks — outperforming Qwen3-30B-A3B-Thinking-2507, Qwen3-32B-Thinking, and even the proprietary Gemini-2.5-Flash-Thinking across multiple benchmarks. Architecture: 48 layers, 15T training tokens, hybrid layout of 12*(3*(Gated DeltaNet->MoE)->(Gated Attention->MoE)). Supports only thinking mode with automatic <think> tag inclusion, may generate longer thinking content.

Fecha de lanzamiento
2025-09-11
Parámetros
80.0B
Longitud del contexto
131K
Modalidades
text

Radar de capacidades

70
general
0
coding
80
reasoning
68
scienceest.
60
agents
0
multimodal

Science se estima a partir de las puntuaciones científicas de LLM Stats o del razonamiento cuando no hay benchmarks científicos dedicados.

Rankings

No hay datos de ranking disponibles

Puntuaciones de benchmarks (LLM Stats)

(LLM Stats (zeroeval))

Chat

IFEvalGoogle Research (2023)88.9%Aut.
Multi-IF77.8%Aut.
TAU-bench Retail69.6%Aut.
Tau2 Airline60.5%Aut.

Code

CFEval2071.00 / 10000Aut.

Communication

Tau2 Retail67.8%Aut.
Tau2 Telecom43.9%Aut.

General

Include78.9%Aut.
BFCL-v372.0%Aut.
Arena-Hard v262.3%Aut.

Language

MMLU-Redux92.5%Aut.
MMLU-Pro82.7%Aut.
MMLU-ProX78.7%Aut.

Math

AIME 202587.8%Aut.
LiveBench 2024112576.6%Aut.
HMMT2573.9%Aut.
PolyMATH56.3%Aut.

Reasoning

GPQANYU + Cohere + Anthropic (2023)77.2%Aut.
LiveCodeBench v668.7%Aut.
SuperGPQA60.8%Aut.
TAU-bench Airline49.0%Aut.
OJBench29.7%Aut.

Writing

WritingBench84.6%Aut.

Índices de evaluación AA

(Artificial Analysis)

No hay datos de evaluación AA disponibles

Puntuaciones por categoría LLM Stats

(LLM Stats (zeroeval))
Instruction Following
80
Language
80
Legal
80
Math
80
Structured Output
80
Finance
80
Biology
80
Chat
70
Physics
70
Reasoning
70
General
70
Healthcare
70
Agents
70
Chemistry
70
Creativity
70
Writing
70
Multimodal
60
Spatial Reasoning
60
Communication
60
Economics
60
Tool Calling
60
Vision
60

Precios

Precio de entrada$0.5 / 1M tokens
Precio de salida$6 / 1M tokens
Precio mixto (3:1)$1.875 / 1M tokens

Velocidad

No hay datos de velocidad disponibles

Ranking de Precios por Proveedor

Ranking de Precios por Proveedor

3 proveedores

Más barato: Alibaba (China)Más caro: Alibaba
ProveedorEntradaSalida
1Alibaba (China)Más barato
$0.144
$1.434
2Qiniu
$0.14493
$1.44928
3AlibabaPRINCIPAL
$0.5
$6

Comparar precios entre diferentes proveedores de API para este modelo.

Fuentes externas