Alibaba Unveils Qwen3.8 Max, a 2.4 Trillion-Parameter AI Model Taking Aim at Kimi K3
Alibaba has entered the latest large language model showdown with Qwen3.8 Max, its biggest AI model so far. The announcement came just two days after Moonshot AI introduced Kimi K3, a massive 2.8 trillion-parameter model, making the timing hard to ignore.
Presented during the World AI Conference in Shanghai, Qwen3.8 Max is being positioned as a major step forward for Alibaba’s AI ambitions. The company claims the model has 2.4 trillion parameters and ranks just behind Anthropic’s Claude Fable 5 in overall capability. However, those claims remain difficult to verify for now.
So far, Alibaba has not released public benchmarks, a model card, licensing details, or technical documentation showing how the model performs in real-world tests. It has also not disclosed how many parameters are active during each request, an important detail for mixture-of-experts models because it affects speed, cost, and hardware requirements.
According to Qwen developer Shuai Bai, Qwen3.8 Max is the first model in the Qwen family above one trillion parameters with multimodal capabilities. That means it is designed to work with text, images, video, and documents, making it more flexible than a traditional text-only chatbot.
Even without public proof of its performance, the model is already available for limited testing. Access is currently offered through Alibaba’s Token Plan subscription service and its developer platforms Qoder and QoderWork. During the preview period, Alibaba says users can test Qwen3.8 Max at one-tenth of the usual price.
Alibaba has also said that open weights are planned, but it has not provided a release date or explained what kind of license will be used. If the company follows through, it would mark a notable shift from its usual approach, as Alibaba has typically kept its largest Max models closed.
For most everyday users, the arrival of Qwen3.8 Max will not change much immediately. A 2.4 trillion-parameter AI model would require enormous computing resources, and even heavily compressed versions would likely need more than a terabyte of storage. That puts it far beyond the reach of nearly all personal computers and consumer devices.
Still, the announcement highlights how intense the competition in artificial intelligence has become. Moonshot AI’s Kimi K3 attracted so much attention that the company reportedly had to pause new subscriptions after its GPU capacity was overwhelmed by demand. Alibaba’s quick response with Qwen3.8 Max shows that China’s AI race is accelerating rapidly, especially among companies building extremely large multimodal models.
The bigger question is whether these trillion-parameter models will deliver practical improvements for users or mainly serve as headline-grabbing demonstrations of scale. Parameter counts can sound impressive, but without independent benchmarks and transparent documentation, it is impossible to know how Qwen3.8 Max truly compares with rival AI systems.
Alibaba’s Qwen lineup has already produced models aimed at more practical use cases. Smaller versions, such as compact models based on the Qwen3.6 family, show that the company is also interested in AI that can run on consumer hardware. That may ultimately matter more to everyday users than massive cloud-based systems that only large companies can afford to operate.
For now, Qwen3.8 Max is best viewed as an ambitious preview rather than a proven breakthrough. Alibaba has made bold claims, but the AI community will be waiting for benchmarks, licensing details, open-weight access, and hands-on testing before judging whether this model truly lives up to its size.
