Skip to main content

Qwen3 Next 80B A3B (Reasoning)

AlibabaQwenOpen WeightApache 2.0 · Commercial OK

Description

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.

Release Date
2025-09-11
Parameters
80.0B
Context Length
131K
Modalities
text

Capability Radar

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

Science is estimated from LLM Stats science scores or reasoning when no dedicated science benchmarks are available.

Rankings

No ranking data available

Benchmark Scores (LLM Stats)

(LLM Stats (zeroeval))

Chat

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

Code

CFEval2071.00 / 10000SR

Communication

Tau2 Retail67.8%SR
Tau2 Telecom43.9%SR

General

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

Language

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

Math

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

Reasoning

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

Writing

WritingBench84.6%SR

AA Evaluation Indices

(Artificial Analysis)

No AA evaluation data available

LLM Stats Category Scores

(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

Pricing

Input Price$0.5 / 1M tokens
Output Price$6 / 1M tokens
Blended Price (3:1)$1.875 / 1M tokens

Speed

No speed data available

Provider Price Ranking

Provider Price Ranking

3 providers

Cheapest: Alibaba (China)Most Expensive: Alibaba
ProviderInputOutput
1Alibaba (China)Cheapest
$0.144
$1.434
2Qiniu
$0.14493
$1.44928
3AlibabaPRIMARY
$0.5
$6

Compare pricing across different API providers for this model.

External Sources