Alibaba has released Qwen3.8-Max, the largest artificial intelligence model in its history — a direct challenge from the open-weight camp to the proprietary frontier systems of OpenAI and Anthropic.
The model, unveiled on Monday, is a 2.4-trillion-parameter mixture-of-experts (MoE) system that activates only about 95 billion parameters per request. Alibaba says the design cuts inference costs while supporting coding, reasoning, and multimodal workloads, and plans to publish open-weight versions next week through Alibaba Cloud's Model Studio alongside a smaller, more deployable Qwen3.8-27B.
In its own benchmarks, Alibaba pits Qwen3.8-Max against Anthropic's Claude Opus 4.8 and Claude Fable 5 and OpenAI's GPT-5.6 Sol on coding tests such as SWE-bench Pro, reporting the highest published score for each rival. The company also claims the model completed a software engineering project from an empty folder in 16 days without human intervention — a headline that analysts say needs far more scrutiny.
Forrester's Charlie Dai called the launch evidence that open-weight models are maturing fast, while Gartner's Nitish Tyagi highlighted the one-million-token context window and sparse architecture as meaningful steps toward cheaper AI-assisted software development. Analysts caution that hosting, licensing, and governance questions remain for enterprises outside China — and that a model card and repository are still needed to confirm the open-weight promise.
