Alibaba released Qwen3.8-Max on Monday, calling it the most capable model in its Qwen family to date. The mixture-of-experts model packs 2.4 trillion total parameters with 95 billion active per query, and is available immediately through QwenCloud — with open weights promised for release next week, the first time Alibaba has open-sourced a model at this scale.

The launch sent Alibaba's stock up 4.5% in pre-market trading in New York and 7% in Hong Kong. Each query can draw on a context window of up to one million tokens — roughly 750,000 words of input — and Alibaba touts capabilities spanning coding, real-world work tasks, long-horizon autonomous operation, and multimodal understanding of documents, video and images.

In demonstrations, the company said the model ran for more than 16 days fully autonomously, accumulating 265 commits, 127 pull requests and 151 issues in a self-built software repository with no human input. In a separate coding competition against 526 human teams under a 24-hour limit, it finished ahead of 87% of the field. The model can also ingest financial reports exceeding 200 pages and footage beyond 100 hours, compiling them into searchable knowledge bases, rebuild software applications from screenshots, and turn 2D architectural floor plans into 3D renderings.

Benchmarks published by Alibaba show Qwen3.8-Max performing at comparable levels to Anthropic's Fable 5 on several coding and general agent tasks, and ahead on some multimodal and document tests: 93.0 versus 88.8 on PaperBench, and 82.8 versus 63.5 on IFBench, though Fable 5 led on SWE-bench Pro and most visual-agent tasks.

The model was previewed at the World AI Conference in Shanghai last month without benchmark data; Monday's release adds a full benchmark table and technical documentation. The same day, Alibaba launched QwenWork, an all-in-one productivity platform in public beta, targeting rivals including Tencent's WorkBuddy and Moonshot AI's Kimi Work. The rollout is part of a wider push backed by a commitment to spend more than 380 billion yuan ($53 billion) on cloud and AI infrastructure over three years.

Sources