Qwen3-Coder-480B-A35B-Instruct Available Now on Novita AI

Alibaba’s groundbreaking Qwen3-Coder-480B-A35B-Instruct is now live on Novita AI.

As one of the most advanced agentic code models available, Qwen3-Coder-480B-A35B-Instruct delivers performance comparable to Claude Sonnet in agentic coding, browser automation, and tool usage.

With 480B parameters and 35B activated, this Mixture-of-Experts model significantly enhances what’s achievable in AI-powered software development and autonomous coding workflows.

Current pricing on Novita AI: $0.95 / M input tokens, $5 / M output tokens

What is Qwen3-Coder-480B-A35B-Instruct?

Qwen3-Coder-480B-A35B-Instruct represents the pinnacle of Alibaba’s coding AI development, featuring advanced agentic coding capabilities. This 480B-parameter Mixture-of-Experts model with 35B activated parameters is specifically designed for autonomous software development, multi-turn programming workflows, and repository-scale understanding.

Technical Specifications

• Type: Causal Language Models
• Total Parameters: 480B with 35B activated
• Architecture: 62 layers, 96 attention heads (Q) and 8 (KV)
• Experts: 160 total with 8 activated
• Context Length: 262,144 tokens natively, extendable to 1M with YaRN
• Training Data: 7.5T tokens with 70% code ratio
• Mode: Non-thinking mode only
• Compatibility: Requires transformers ≥4.51.0

Key Features and Performance

Core Capabilities

Advanced Agentic Coding: Significant improvements in autonomous code generation, debugging, and software engineering workflows that compete with premium commercial models.

Long-Context Understanding: Native 256K token support with 1M token extension enables comprehensive repository analysis and cross-file reasoning.

Tool Integration: Specialized function calling format supports popular platforms including Qwen Code, CLINE, Claude Code, and Trae.

Multi-Turn Development: Strong performance in complex tasks requiring planning, tool usage, feedback processing, and iterative development cycles.

Performance Highlights

Qwen3-Coder-480B-A35B-Instruct achieves strong results among open-source models across multiple benchmarks:

• Agentic Coding: Advanced autonomous programming capabilities
• Browser Automation: Impressive results in web automation workflows
• SWE-Bench: Notable performance on real-world software engineering tasks
• Multi-Language Support: Solid performance across programming languages and frameworks

How to Access Qwen3-Coder-480B-A35B-Instruct on Novita AI

Playground Interface (No Coding Required)

• Quick Setup: Sign up and start experimenting with Qwen3-Coder-480B-A35B-Instruct alongside other top models
• Interactive Testing: Test complex coding prompts and evaluate agentic capabilities in real-time
• Model Comparison: Easily compare Qwen3-Coder-480B-A35B-Instruct with other leading coding models

API Integration (For Developers)

Connect to your applications using Novita AI’s unified REST API:

from openai import OpenAI
  
client = OpenAI(
    base_url="https://api.novita.ai/v3/openai",
    api_key="",
)

model = "qwen/qwen3-coder-480b-a35b-instruct"
stream = True # or False
max_tokens = 131072
system_content = ""Be a helpful assistant""
temperature = 1
top_p = 1
min_p = 0
top_k = 50
presence_penalty = 0
frequency_penalty = 0
repetition_penalty = 1
response_format = { "type": "text" }

chat_completion_res = client.chat.completions.create(
    model=model,
    messages=[
        {
            "role": "system",
            "content": system_content,
        },
        {
            "role": "user",
            "content": "Hi there!",
        }
    ],
    stream=stream,
    max_tokens=max_tokens,
    temperature=temperature,
    top_p=top_p,
    presence_penalty=presence_penalty,
    frequency_penalty=frequency_penalty,
    response_format=response_format,
    extra_body={
      "top_k": top_k,
      "repetition_penalty": repetition_penalty,
      "min_p": min_p
    }
  )

if stream:
    for chunk in chat_completion_res:
        print(chunk.choices[0].delta.content or "", end="")
else:
    print(chat_completion_res.choices[0].message.content)
  
  

Code with Qwen3-Coder

Qwen Code – Specialized CLI Tool

Qwen Code is a research-purpose CLI tool adapted from Gemini CLI, with enhanced parser and tool support for Qwen-Coder models.

Prerequisites: Ensure you have Node.js version 20 or higher installed. You can download it from the official Node.js website.

Install the package globally:

bash

npm install -g @qwen-code/qwen-code

Step 3: Configure Environment Variables

For Windows (Command Prompt):

set OPENAI_API_KEY=Your_Novita_API_Key

set OPENAI_BASE_URL=https://api.novita.ai/v3/openai

set OPENAI_MODEL=qwen/qwen3-coder-480b-a35b-instruct

For Linux and Mac (Bash):

export OPENAI_API_KEY=”Your_Novita_API_Key”

export OPENAI_BASE_URL=”https://api.novita.ai/v3/openai”

export OPENAI_MODEL=”qwen/qwen3-coder-480b-a35b-instruct”

With installation and configuration complete, you can now start Qwen Code in your project directory. Navigate to your desired project location using the cd command:

cd <your-project-directory>
qwen .

Cline – VSCode Integration

Setup Steps (See detailed setup guide):

1. Install Cline extension in VSCode (Ctrl + Shift + X, search “Cline”)
2. Click Cline icon in sidebar
3. Select “OpenAI Compatible” provider
4. Configure:
   • Base URL: https://api.novita.ai/v3/openai
   • API Key: Your Novita AI API key
   • Model: qwen/qwen3-coder-480b-a35b-instruct-fp8

Dual-Mode Workflow:

• Plan Mode: Strategic analysis and planning with full repository understanding
• Act Mode: Implementation execution with direct code changes and file management

Trae – Native Integration

Trae provides built-in Novita AI integration – simply select Qwen3-Coder-480B-A35B-Instruct from Trae’s model interface for immediate access to agentic coding capabilities.

Check detailed setup guide for Trae.

Best Practices

Recommended Parameters

• Temperature: 0.7
• Top P: 0.8
• Top K: 20
• Repetition Penalty: 1.05

Output Configuration

• Standard Tasks: 65,536 tokens for comprehensive code generation
• Repository-Scale: Leverage full 256K context for large codebase analysis

Conclusion

Qwen3-Coder-480B-A35B-Instruct represents a significant advancement in autonomous coding AI, delivering performance comparable to Claude Sonnet while maintaining open-source accessibility. With advanced agentic capabilities, extensive tool integration, and repository-scale understanding, this model significantly enhances how developers approach complex software engineering tasks.

The model’s smooth integration with popular development tools like Qwen Code, Claude Code, Cline, and Trae makes it a valuable addition to modern development workflows, enabling developers to tackle complex challenges with AI assistance.

Ready to enhance your coding workflow? Try Qwen3-Coder-480B-A35B-Instruct on Novita AI’s platform today.

---

Novita AI is an AI cloud platform that offers developers an easy way to deploy AI models using our simple API, while also providing affordable and reliable GPU cloud for building and scaling AI applications.