On August 3, 2026, Alibaba announced Qwen3.8-Max, its largest and most powerful AI model yet. The system contains 2.4 trillion parameters, and Alibaba says it can match or beat some of the best AI models currently available. The model is accessible through Alibaba Cloud's Model Studio platform, and the company plans to release the model's weights — the underlying data that lets others run it on their own computers — the following week (Reuters, Bloomberg).

Think of parameters as the connections inside an AI model — the more there are, the more complex the patterns the model can learn. A model with 2.4 trillion of these connections can handle a wide range of tasks, from writing code to analyzing documents.

The model is built on the foundation of Qwen 3.5, Alibaba's earlier system, according to the company's Qwen site. The announcement post, titled "Qwen3.8-Max: A New Bar for Coding and Cowork," claims improvements across coding, work, and research.

Alibaba's own testing shows Qwen3.8-Max broadly matches, and sometimes exceeds, the performance of Fable 5, Anthropic's top model, on benchmark tests (The Verge. The company also says the model rivals the best systems from OpenAI and from Moonshot AI's Kimi K3, a Chinese competitor. On the Arena.AI leaderboard — a website where people compare AI models by voting on which gives better answers — Qwen3.8-Max ranks below only Fable 5 and three models in Anthropic's Opus family. Bloomberg separately reports that Qwen3.8-Max ranks higher than Kimi K3 on several benchmarks.

Qwen3.8-Max is slightly smaller than Moonshot AI's Kimi K3, which has 2.8 trillion parameters (Bloomberg). Alibaba had previewed Qwen3.8-Max the previous month, calling it "second only to Fable 5."

The Qwen team also published a tutorial blog post showing how to build a tool that generates article cover images using the model. Alibaba's Qwen Studio platform offers chatbot interaction, image and video understanding, image generation, and document processing (Qwen blog).

The decision to release the model's weights within a week of the announcement is unusual for a system this large. Most top-tier AI models from companies like Anthropic and OpenAI are only available through paid online services. If Alibaba follows through, developers and researchers would be able to run Qwen3.8-Max on their own machines, which could change how academic and business teams experiment with and deploy AI.

That said, Alibaba's benchmark claims warrant caution. The comparisons to Fable 5 come from Alibaba's own testing, and the Arena.AI leaderboard reflects user preferences rather than standardized exams. Independent testing will be the real test once the weights are available.

The comparison to Kimi K3 is also telling. Kimi K3 is larger, yet Qwen3.8-Max outperforms it on several benchmarks according to both Alibaba and Bloomberg. This suggests that bigger does not always mean better when it comes to AI models. For anyone choosing between models, the practical takeaway is that Alibaba is competing on two fronts: performance that rivals the best, and open access that closed competitors do not offer. Whether Qwen3.8-Max holds up under independent testing will determine whether it is a genuine turning point or another strong entry in a fast-moving field.