Access the world's fastest-improving
Qwen model family
47 Qwen models from 0.8B to 480B parameters — language, code, vision, reasoning, and TTS. Zero platform fee. Multi-provider failover. One unified API.
Available Models
Chat, code, vision, reasoning, image & video generation — all through one platform.
Qwen3.6 Plus
Best balance of speed and quality. #1 most-used model on OpenRouter.
Qwen3 Coder Plus
Top coding model. Generation, completion, and refactoring across 80+ languages.
Qwen Image 2.0 Pro
High-quality text-to-image generation with strong text rendering.
Wan 2.7 T2V
Text-to-video generation with audio support. 2-15 seconds at up to 1080P.
See It in Action
Watch videos generated through Qwen models on MagicRouter.
Qwen Models Are Top-Ranked Globally
Independent benchmarks confirm Qwen as the leading open-weight model family.
| Benchmark | Qwen3.5 397B | GPT-4.1 | Claude Opus 4.6 | Gemini 2.5 Pro |
|---|---|---|---|---|
| MMLU | 88.5 | 90.2 | 89.5 | 89.8 |
| GPQA Diamond | 88.4 #1 open | 66.3 | 88.5 | 84.0 |
| AIME 2025 | 81.5 | — | 99.8 | 95.0 |
| SWE-bench Verified | 76.4 | 54.6 | 80.8 | — |
| IFBench | 76.5 #1 | — | 58.0 | — |
| BFCL-V4 (Tool Use) | 72.9 #1 | — | 63.3 | — |
Qwen3.5-397B scores 88.4 — the highest GPQA Diamond score from any open-weight model
Leads BFCL-V4 at 72.9 and IFBench at 76.5, beating both closed and open models
Qwen3.5-9B beats OpenAI's GPT-OSS-120B on MMLU-Pro (82.5 vs 80.8)
Qwen3.5 leads in tool use and instruction following. Claude Opus 4.6 leads in math reasoning (AIME) and software engineering (SWE-bench). GPT-4.1 leads in MMLU knowledge.
Sources: BenchLM.ai · LMSYS Chatbot Arena · LLM Stats · VentureBeat · Qwen Blog · Claude Benchmarks · OpenAI GPT-4.1
Qwen 3.6 Plus Is the #1 Most-Used Model
Live usage rankings from OpenRouter — the largest multi-model API gateway processing 20T+ tokens/week.
| Rank | Model | Provider | Weekly Tokens | Trend |
|---|---|---|---|---|
| 1 | Qwen3.6 Plus | Alibaba | 5.53T | NEW |
| 2 | MiMo V2 Pro | Xiaomi | 2.41T | +42% |
| 3 | Qwen3.6 Plus Preview | Alibaba | 1.62T | +9,088% |
| 4 | Step 3.5 Flash | Stepfun | 1.25T | +15% |
| 5 | DeepSeek V3.2 | DeepSeek | 1.19T | +4% |
| 6 | MiniMax M2.7 | MiniMax | 1.18T | +10% |
| 7 | Claude Sonnet 4.6 | Anthropic | 1.03T | +2% |
| 8 | Claude Opus 4.6 | Anthropic | 1.0T | +2% |
| 9 | Gemini 3 Flash | 996B | +2% | |
| 10 | MiniMax M2.5 | MiniMax | 918B | +10% |
Qwen3.6 Plus took #1 within days of launch — 5.53T tokens/week
Most downloaded open-source LLM family, surpassing Meta Llama
Deployed via Alibaba Cloud Model Studio for production workloads
Source: OpenRouter Live Rankings · Digital Applied Analysis · Open Source For You — Rankings reflect rolling 7-day token volume as of April 2026.
Why Qwen on MagicRouter?
Get the best of open-weight AI with zero platform fees and smart routing.
on all 47 Qwen models
Client-side routing, no proxy
Alibaba, Together, DeepInfra & more
Always the cheapest healthy provider
Build with the best open-weight models. Zero fees.
Free tier available. No credit card required.