inference

Inference

Qwen3 Coder 480B A35B

Text

Model overview

Price

$1.00 - $1.50

Input - Output

Parameters

35B - 480B

Active - Total

Context window

262K

Release date

Jul 2025

Qwen3-Coder-480B-A35B-Instruct is a Mixture-of-Experts (MoE) code generation model developed by the Qwen team. It is optimized for agentic coding tasks such as function calling, tool use, and long-context reasoning over repositories. The model features 480 billion total parameters, with 35 billion active per forward pass (8 out of 160 experts).

Alibaba

License: apache-2.0

Qwen/Qwen3-Coder-480B-A35B-Instruct

Supported features

Reasoning

JSON mode

Structured output

Tool calling

Use this model

import openai
import weave
# Weave autopatches OpenAI to log LLM calls to W&B
weave.init("<team>/<project>")
client = openai.OpenAI(
    # The custom base URL points to W&B Inference
    base_url='https://api.inference.wandb.ai/v1',
    # Get your API key from https://wandb.ai/authorize
    # Consider setting it in the environment as OPENAI_API_KEY instead for safety
    api_key="<your-apikey>",
    # Optional: Team and project for usage tracking
    project="<team>/<project>",
)
response = client.chat.completions.create(
    model="Qwen/Qwen3-Coder-480B-A35B-Instruct",
    messages=[
        {"role": "system", "content": "You are a helpful assistant."},
        {"role": "user", "content": "Tell me a joke."}
    ],
)
print(response.choices[0].message.content)

Enter to Rename, Shift+Enter to Preview