Alibaba has launched Qwen 3.6-Max-Preview, its most advanced AI model to date. This new model surpasses its predecessor, Qwen 3.6-Plus, demonstrating significant improvements in world knowledge and instruction following, and leading in six major coding and agentic benchmarks.
The model is available via Qwen Studio and the Alibaba Cloud Model Studio API. Unlike previous open-source offerings, Qwen 3.6-Max-Preview is a proprietary, hosted model without open weights. Its API is designed for seamless integration, compatible with both OpenAI and Anthropic specifications, allowing developers to plug it into existing systems with minimal changes.
This strategic shift aligns with a broader trend among Chinese AI labs, moving from free open-source models to monetized, proprietary services. Alibaba's Qwen series has seen rapid adoption, contributing significantly to China's growing influence in the global AI landscape.
Qwen 3.6-Max-Preview has claimed top positions in benchmarks like SWE-bench Pro for real-world software engineering and Terminal-Bench 2.0 for command-line execution. It also shows strong performance in general problem-solving, tool use, web interaction, and scientific programming.
Key features include an improved agentic coding capability and enhanced real-world agent and knowledge reliability. The model supports a 256k token context window and is recommended for agentic tasks requiring continuity, such as autonomous agents or long-running code generation workflows.