Qwen3-Next-80B-A3B-Instruct

Next-generation instruction model with extreme efficiency

About model

Instruction-Optimized Efficiency:
Qwen3-Next Instruct features a highly sparse MoE structure that activates only 3B of its 80B parameters during inference. Supports only instruct mode without thinking blocks, delivering performance on par with Qwen3-235B-A22B-Instruct-2507 on certain benchmarks while using less than 10% training cost and providing 10x+ higher throughput on contexts over 32K tokens.

‍

Quickstart guides

RAG

Building a RAG Workflow

Agents

Agent Workflows

Apps

Next.js Chat Quickstart

API usage

cURL
Python
Typescript

Endpoint:

Qwen/Qwen3-Next-80B-A3B-Instruct

curl -X POST "https://api.together.xyz/v1/chat/completions" \
  -H "Authorization: Bearer $TOGETHER_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "Qwen/Qwen3-Next-80B-A3B-Instruct",
    "messages": [
      {
        "role": "user",
        "content": "What are some fun things to do in New York?"
      }
    ]
}'

from together import Together

client = Together()

response = client.chat.completions.create(
  model="Qwen/Qwen3-Next-80B-A3B-Instruct",
  messages=[
    {
      "role": "user",
      "content": "What are some fun things to do in New York?"
    }
  ]
)
print(response.choices[0].message.content)

import Together from 'together-ai';
const together = new Together();

const completion = await together.chat.completions.create({
  model: 'Qwen/Qwen3-Next-80B-A3B-Instruct',
  messages: [
    {
      role: 'user',
      content: 'What are some fun things to do in New York?'
     }
  ],
});

console.log(completion.choices[0].message.content);

Model card
Architecture Overview:
• 48 layers with 2048 hidden dimension and hybrid layout pattern
• 512 total experts with 10 activated and 1 shared expert per MoE layer
• Multi-token prediction mechanism for faster inference

Instruction Optimization:
• Supports only instruct mode without blocks
• Specialized post-training for task completion on 15T tokens
• Performance on par with Qwen3-235B while using significantly fewer resources

Performance Characteristics:
• 262K native context length, extensible to 1M tokens with YaRN scaling
• More than 10x higher throughput on contexts over 32K tokens
• SGLang and vLLM deployment support with Multi-Token Prediction
‍
Applications & use cases
Task Automation:
• Code generation and software development assistance with cost-effective processing
• Content creation and editing with specific instructions
• Data analysis and report generation following detailed guidelines

Business Applications:
• Customer service automation with instruction-based responses
• Technical documentation generation with specific formatting requirements
• Process automation and workflow optimization

Agentic Use Cases:
• Tool calling capabilities with MCP configuration support
• Multi-step task execution with built-in and custom tools
• Extended conversation and context-aware task completion up to 262K tokens
‍

Related models

Model specifications

Model data

Model provider
Qwen
Type
Chat
LLM
Main use cases
Chat
Function Calling
Features
Function Calling
Speed
Very High
Intelligence
Medium
Deployment
Serverless
On-Demand Dedicated
Monthly Reserved
Endpoint
Qwen/Qwen3-Next-80B-A3B-Instruct
Parameters
81.3B
Context length
256K
Input price
$0.15 / 1M tokens
Output price
$1.50 / 1M tokens
Input modalities
Text
Output modalities
Text

Released
September 9, 2025
Last updated
September 12, 2025
Quantization level
BF16
External link
Provider docs
Category
Chat

Run in Playground

Quickstart docs

Deploy model

Qwen3-Next-80B-A3B-Instruct

About model

API usage

Model card

Applications & use cases