Alibaba's Breakthrough 80B-Parameter Open-Source AI
Experience unprecedented efficiency with Qwen3-Next: 80B parameters, only 3B active per inference. Ultra-long 262K context, multilingual excellence, and enterprise-ready deployment.
Unparalleled Efficiency and Scale
Revolutionary sparse MoE architecture that delivers cutting-edge performance while slashing compute costs by up to 90%
Extreme Efficiency
Activate only 3B of 80B parameters per token prediction. Achieve 90% cost savings with breakthrough sparse MoE design.
Ultra-Long Context
Process 262K tokens in a single pass with hybrid DeltaNet and gated attention. Perfect for analyzing entire books or contracts.
Multi-Token Prediction
Generate multiple tokens per step, reducing latency and enhancing user experiences for real-time applications.
Enterprise Ready
Apache 2.0 licensed with on-premises, VPC, and private cloud deployment options for maximum security and compliance.
Multilingual Excellence
Native support for 119 languages with seamless cross-lingual capabilities for global applications.
Agentic Tool Use
Integrate with APIs, databases, and external tools for complex multi-step workflows and autonomous reasoning.
State-of-the-Art Performance
Qwen3-Next consistently outperforms leading models while maintaining extreme efficiency
| Benchmark | Qwen3-Next-80B | Qwen3-32B | Llama 3.1-70B |
|---|---|---|---|
| MMLU | 78.5% | 76.2% | 77.8% |
| HumanEval | 82.1% | 79.5% | 81.2% |
| GSM8K | 91.2% | 88.7% | 90.1% |
| MT-Bench | 84.3% | 81.9% | 83.5% |
Get Started in Minutes
Integrate Qwen3-Next into your applications with simple API calls
import os
from dashscope import Generation
# Set your API key
os.environ['DASHSCOPE_API_KEY'] = 'your_api_key'
# Call Qwen3-Next with 80B params, 3B active
response = Generation.call(
model='qwen3-next-80b-a3b-instruct',
prompt='Explain the benefits of sparse MoE architectures.',
max_tokens=200,
temperature=0.7
)
if response.status_code == 200:
print(response.output['text'])
else:
print(f"Error: {response.message}")
curl -X POST 'https://dashscope.aliyuncs.com/api/v1/services/aigc/text-generation/generation' \
-H 'Authorization: Bearer YOUR_API_KEY' \
-H 'Content-Type: application/json' \
-d '{
"model": "qwen3-next-80b-a3b-instruct",
"input": {
"prompt": "Explain the benefits of sparse MoE architectures."
},
"parameters": {
"max_tokens": 200,
"temperature": 0.7
}
}'
import { Generation } from '@alicloud/dashscope';
const client = new Generation({
apiKey: process.env.DASHSCOPE_API_KEY
});
async function callQwen3Next() {
try {
const response = await client.call({
model: 'qwen3-next-80b-a3b-instruct',
prompt: 'Explain the benefits of sparse MoE architectures.',
max_tokens: 200,
temperature: 0.7
});
console.log(response.output.text);
} catch (error) {
console.error('Error:', error.message);
}
}
callQwen3Next();
Get API Key
Sign up for Alibaba Cloud and obtain your DashScope API key
Install SDK
Use pip, npm, or curl to integrate with your preferred stack
Start Building
Make API calls and experience the power of 80B sparse parameters
Build Smarter Applications
From long-document analysis to agentic workflows, Qwen3-Next powers next-generation AI applications
Long-Document Understanding
Process entire books, legal contracts, or research papers in a single 262K context window. Eliminate chunking limitations and maintain global coherence.
Agentic Tool Use
Build autonomous agents that integrate with APIs, databases, and external tools for complex multi-step workflows and reasoning tasks.
Enterprise Compliance
Deploy securely in your infrastructure to meet governance, audit, and data privacy requirements with full control over your AI workloads.
Ready to Experience the Future of AI?
Join thousands of developers building with Qwen3-Next. Start your journey with 80B parameters and 3B active efficiency.
✨ Free tier includes 10,000 API calls per month. No credit card required.