rewrite this title Alibaba Drops Qwen 3.6 Max Preview—Its Most Powerful Model Yet – Decrypt

Alibaba released a preview of its next flagship AI model on Monday, pushing the Qwen series further into the frontier race. Qwen 3.6-Max-Preview is the most powerful model the company has shipped so far, topping six major coding benchmarks and posting meaningful gains in world knowledge and instruction following over its predecessor, Qwen 3.6-Plus.

The model is available now through Qwen Studio and the Alibaba Cloud Model Studio API under the string qwen3.6-max-preview. It is a proprietary, hosted model with no open weights, and its API is compatible with both OpenAI and Anthropic specifications, meaning developers can plug it into existing pipelines with minimal changes.

This is also a shift in Alibaba’s business model, as it was known to provide powerful open-source models by default. The lower end models are still open source.

According to Qwen’s official blog, Qwen 3.6-Max-Preview ranked first across several major benchmarks that test coding and agent capabilities, including SWE-bench Pro (real-world software engineering tasks), Terminal-Bench 2.0 (command-line execution), SkillsBench (general problem-solving), QwenClawBench (tool use), QwenWebBench (web interaction), and SciCode (scientific programming).

In terms of gains over Qwen 3.6-Plus, benchmarks for agentic skills put it on top of the family and other models like Calude 4.5 or GLM 5.1. The model also improved in knowledge benchmarks, with SuperGPQA (advanced reasoning) increasing by 2.3% and QwenChineseBench (Chinese language performance) by 5.3%. Instruction-following ability, measured by ToolcallFormatIFBench, put it on top of the rankings, beating Claude.

The release lands three days after Alibaba open-sourced Qwen 3.6-35B-A3B, a 35-billion-parameter model that activates only 3 billion parameters per inference.

That approach is designed to cut compute costs without sacrificing output quality. Combined with Monday’s Max release, the Qwen 3.6 lineup now spans Max-Preview at the top of the family, the Qwen Plus variance for balanced workloads, Flash for speed-first tasks, and 35B-A3B for those running locally.

The Max-Preview also ships with preserve_thinking, a feature that carries reasoning traces across multi-turn conversations. Alibaba specifically recommends it for agentic tasks where context continuity matters. For developers running autonomous agents or long-running code generation workflows, that is a meaningful addition.



As Decrypt reported last week, Alibaba shut down the free tier of Qwen Code just days after fellow Chinese lab MiniMax rewrote its open-source license to block commercial use without written authorization. Both moves signal a broader shift: Chinese AI labs that built massive adoption on free, open services are now pivoting toward monetized, proprietary offerings. Qwen overtook Meta’s Llama as the most deployed self-hosted model on the planet—and that momentum was built almost entirely on free access.

That free-to-paid transition runs parallel to another trend. Chinese open models went from 1.2% of global open-model usage in late 2024 to roughly 30% by end of 2025, with Qwen leading the charge. The Qwen 3.6-Max-Preview is the proprietary tip of that spear as it’s the model Alibaba is betting will compete directly with OpenAI’s GPT and Anthropic’s Claude at the frontier.

Qwen 3.6-Max-Preview is explicitly labeled a work in progress. Alibaba said the model is still under active development and expects further gains in future versions. Independent benchmarking from Artificial Analysis puts it as the second best performing model behind Muse Spark—well above the median of comparable reasoning models in its price tier. The model supports a 256k token context window and handles text only, with no image input at launch.