Qwen3.6 Tops Artificial Analysis as the Best Chinese Model
🏆 Alibaba's Qwen3.6-Max-Preview Tops the Charts: Artificial Analysis Crowns It the Best Chinese Model
Alibaba has released Qwen3.6-Max-Preview, and the respected independent benchmark Artificial Analysis shows:
Qwen3.6-Max-Preview outperforms GLM-5.1, MiniMax-M2.7, and more — taking the crown as the best Chinese model!
📊 Artificial Analysis Leaderboard Results
Artificial Analysis is a respected independent AI model evaluator, and its Intelligence Index v4.0 covers 10 evaluation dimensions:
| Benchmark | What it measures |
|---|---|
| GDPval-AA | General problem solving |
| τ²-Bench Telecom | Telecom-domain tasks |
| Terminal-Bench Hard | Terminal command tasks |
| SciCode | Scientific coding |
| AA-LCR | Long-context reasoning |
| AA-Omniscience | Knowledge reliability |
| IFBench | Instruction following |
| Humanity’s Last Exam | The ultimate exam |
| GPQA Diamond | Graduate-level questions |
| CritPt | Critical thinking |
🎯 Beating Its Chinese Rivals
| Model | Provider | Rank |
|---|---|---|
| Qwen3.6-Max-Preview | Alibaba | 🥇 First |
| GLM-5.1 | Zhipu AI | 🥈 Second |
| MiniMax-M2.7 | MiniMax | 🥉 Third |
| DeepSeek V3 | DeepSeek | Fourth |
Qwen3.6-Max-Preview in the overall evaluation beat every Chinese rival and claimed the top spot as the best Chinese model.
✨ Qwen3.6-Max-Preview Features
1. Million-Token Context
Qwen3.6-Max-Preview supports a 1M-token context window:
- Long documents: handles ultra-long documents and codebases
- Multi-turn chats: keeps context across long conversations
- Code comprehension: understands large code projects
2. Strong Coding
The Qwen family has long been known for its coding chops:
- Code generation: high-quality code generation
- Code explanation: understands complex code logic
- Bug fixing: pinpoints and fixes issues precisely
- Multi-language support: Python, Java, C++, Go, and more
3. Great Value for Money
Qwen3.6 series pricing is extremely competitive:
| Model | Input price | Output price |
|---|---|---|
| Qwen3.6-Plus | $2/1M tokens | $8/1M tokens |
| Claude Sonnet 4.6 | $3/1M tokens | $15/1M tokens |
| GPT-5.4 | $2.5/1M tokens | $10/1M tokens |
Qwen is the cheaper option, yet it gives up nothing to top international models on performance.
📈 Intelligence vs. Price
Artificial Analysis' Intelligence vs. Price chart highlights the best value-for-money zone:
- Qwen3.6-Max-Preview sits in the "most attractive quadrant"
- High intelligence index + Low price = best value for money
- Beats MiniMax and Zhipu GLM and other Chinese rivals
💡 AA-Omniscience: Knowledge Reliability
AA-Omniscience Index measures a model's knowledge reliability and hallucination rate:
- Rewards correct answers
- Penalizes hallucinations
- Doesn't penalize refusing to answer
- Score range: -100 to 100
Qwen3.6-Max-Preview performed excellently on the Omniscience test — low hallucination rate, reliable knowledge.
🚀 How to Use It
API access
# 通义千问 API
curl -X POST https://dashscope.aliyuncs.com/api/v1/services/aigc/text-generation/generation \
-H "Authorization: Bearer $API_KEY" \
-H "Content-Type: application/json" \
-d '{
"model": "qwen3.6-max-preview",
"input": {"prompt": "Hello, Qwen!"},
"parameters": {}
}'
Open-source versions
The Qwen3.6 series has open-source versions:
- Qwen3.6-7B: lightweight open-source version
- Qwen3.6-72B: high-performance open-source version
- Apache 2.0 license: free for commercial use
OpenClaw support
Qwen models are already integrated into OpenClaw:
# OpenClaw 配置 openclaw config set model qwen-portal/qwen3.6-max-preview
📝 Wrap-Up
Alibaba's Qwen3.6-Max-Preview tops the Artificial Analysis ranking of Chinese models:
- 🥇 No. 1 among Chinese models on the Intelligence Index
- 🏆 Ahead of GLM-5.1 and MiniMax-M2.7
- 📚 Million-token context, strong coding
- 💰 Great value: low price, high performance
- ✅ High knowledge reliability on AA-Omniscience
- 🚀 Both API and open-source options
In the race among Chinese LLMs, Alibaba's Qwen3.6 is out in front for now.
Artificial Analysis:https://artificialanalysis.ai/models
Tongyi Qianwen official site:https://tongyi.aliyun.com
Qwen GitHub:https://github.com/QwenLM