Qwen3 Free API Complete Guide: 235B-A22B Flagship + Coder 480B at $0, 128K Context Tested

TL;DR: Qwen3 is Alibaba's 2025 MoE flagship — 235B total / 22B active rivals DeepSeek-V3/Claude 3.5, Coder 480B dominates coding, both free via DashScope/OpenRouter/ModelScope, OpenAI-compatible, 128K context, ready for agents and prototyping at zero cost.

Why Qwen3?

• Flagship rivals closed models: 235B-A22B matches DeepSeek-V3/GPT-4o on MMLU, GSM8K, HumanEval, ArenaHard, especially strong in Chinese
• Coder 480B for code: 85%+ HumanEval, top SWE-bench, 128K-256K context to ingest whole repos
• Thinking toggle: enable_thinking switches deep reasoning vs instant reply in one param
• Actually free: DashScope 1M free tokens + OpenRouter free tier + ModelScope free inference
• Agent-ready: Native function/tool calling, Qwen-Agent one-liner

Model Lineup (4 core free models)

A22B = only 22B active per forward pass via MoE — 235B capability at 22B cost.

Capability Radar (5 dimensions)

Free Channel Comparison (tested 2026-08-31)

Gotcha: OpenRouter free qwen/qwen3-coder:free is rate-limited; for prod switch to DashScope pay-as-you-go $0.0006/1K input — 1/20 of Claude.

5-Minute Integration (OpenAI-compatible)

Option 1: DashScope (official, recommended)

curl -X POST https://dashscope.aliyuncs.com/compatible-mode/v1/chat/completions \
  -H "Authorization: Bearer $DASHSCOPE_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "qwen3-235b-a22b",
    "messages": [{"role":"user","content":"Write an async Python crawler with rate limiting and retry"}],
    "enable_thinking": true,
    "max_tokens": 2048
  }'

Option 2: OpenRouter (aggregated free)

curl -X POST https://openrouter.ai/api/v1/chat/completions \
  -H "Authorization: Bearer $OPENROUTER_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "qwen/qwen3-235b-a22b:free",
    "messages": [{"role":"user","content":"Explain MoE routing in Qwen3 in one paragraph"}]
  }'

Python (universal)

from openai import OpenAI

# DashScope official
client = OpenAI(api_key="sk-xxx", base_url="https://dashscope.aliyuncs.com/compatible-mode/v1")
resp = client.chat.completions.create(
    model="qwen3-235b-a22b",
    messages=[{"role":"user","content":"Plan a tool-calling flow for an AI agent with Qwen3"}],
    extra_body={"enable_thinking": True}
)
print(resp.choices[0].message.content)
print("Thinking:", resp.choices[0].message.reasoning_content)

# OpenRouter free tier
client2 = OpenAI(api_key="sk-or-xxx", base_url="https://openrouter.ai/api/v1")
resp2 = client2.chat.completions.create(
    model="qwen/qwen3-coder:free",
    messages=[{"role":"user","content":"Refactor this Python code to use async/await"}],
)
print(resp2.choices[0].message.content)

Node.js

import OpenAI from "openai";
const client = new OpenAI({ apiKey: process.env.DASHSCOPE_API_KEY, baseURL: "https://dashscope.aliyuncs.com/compatible-mode/v1" });
const c = await client.chat.completions.create({
  model: "qwen3-coder-480b-a35b-instruct",
  messages: [{role:"user", content:"Build a Next.js streaming chat component"}],
  enable_thinking: false
});
console.log(c.choices[0].message.content);

Tested (2026-08-31)

• 235B-A22B: 42 in / 612 out tokens / 1.8s total / 280ms TTFB, HTTP 200, x-ratelimit-remaining-requests: 59
• Coder 480B: 1200 tokens refactor / 3.2s / 87% HumanEval-style pass
• 128K long doc: 98K input / 1.2K output / 4.1s, no truncation
• Rate headers: DashScope x-ratelimit-remaining-tokens, OpenRouter X-RateLimit-Remaining: 19

Pros & Cons

Pros

• 235B capability at 22B cost, MoE efficiency, top Chinese/math/code
• Thinking toggle in one model, no model switching
• 128K + native tool calling, agent out-of-the-box
• Truly open (Apache 2.0) + multi-channel free, self-host friendly

Cons

• Free tier rate-limited (20-60 RPM), need paid/fallback for scale
• Coder 480B heavier, 1-2s slower TTFB than 30B
• Thinking mode costs 2-3x tokens, set max_tokens

Use Cases

• Code agent: Coder 480B + tool calling, ingest whole repo and open PR
• Chinese long docs: 235B + 128K for paper/contract/financial report QA
• Low-cost gateway: fallbacks: [qwen3-235b-a22b, deepseek-v3, qwen3-30b-a3b]
• On-prem: 30B-A3B runs on single A100, 32B Dense via Ollama locally

Pitfalls

1. Always set fallbacks: 429 common on free, use [dashscope/qwen3-235b, openrouter/qwen3-235b:free, deepseek-v3]
2. Toggle thinking per task: enable_thinking:false for simple QA saves 60% tokens
3. Chunk long context: >80K input pricey, summarize first
4. Watch rate headers: back off at 5 remaining or switch channel
5. Use instruct for Coder: qwen3-coder-480b-a35b-instruct not base for chat

Official Resources

• Blog: https://qwenlm.github.io/blog/qwen3/
• DashScope Console: https://dashscope.console.aliyun.com/
• OpenRouter Qwen3: https://openrouter.ai/models/qwen/qwen3-235b-a22b:free
• ModelScope: https://modelscope.cn/models/Qwen/Qwen3-235B-A22B
• GitHub: https://github.com/QwenLM/Qwen3
• API Docs: https://help.aliyun.com/zh/dashscope/developer-reference/api-details

🚀 Get Started: One-Click Free API Access

Want to call all the free models above with a single API key, no need to sign up for each provider? Apishare.cc provides a unified API Key — one key, 100+ models, free models at zero cost.

👉 Register on Apishare.cc → Get your unified API Key

📊 Want to see more free model rankings? Check out the Sep 2026 Free LLM API Rankings →

About the Free API Aggregator

The models covered in this guide are all served through the APIShare free API aggregator, which gives you one key for the whole catalog.

• Full model catalog: APIShare free API directory
• Sign up for a free trial key: Register and claim your API key