Ali’s new flagship model Qwen3.8 preview: 24 trillion parameters, rivaling Fable 5 and Kimi 3

Alibaba has officially joined the global parameter arms race for top-tier large models, bringing the "fire of battle" into its own investment territory.

On July 19, Alibaba released the flagship large model Qwen3.8 Max Preview, with a parameter scale as high as 2.4 trillion. Officially, it claims its overall capability ranks second globally, only after Anthropic's Fable 5. At the same time, Alibaba has promised to "soon" release the full model weights.

Notably, the timing of this release is intriguing—just days earlier, Moonshot AI, a startup invested in by Alibaba, stirred global tech stocks with its Kimi K3 with 2.8 trillion parameters, and briefly broke the record for the "largest open model." The high-profile launch of Qwen3.8 Max has intensified the competition within Alibaba's investment landscape for the "largest open model."

Currently, the preview version is online at Alibaba's code development platforms Qoder, QoderWork, and Token Plan, and the public dialogue version is open simultaneously.

Preview Version Released Early, Weight Release Timing Undetermined

Alibaba announced the news through its official announcement and social media posts from the Tongyi Qianwen team, positioning Qwen3.8 Max as "one of the most powerful models today, comparable to leading frontier AI models."

Developers can now try out the preview version in advance via platforms such as Qoder, and the public online conversation portal is also open. Alibaba stated it will "soon" release the full open weights, but did not disclose a specific date.

According to reports, developer Bai Shuai, who participated in the project, said that Qwen3.8 Max is the team's first multimodal model with parameter scale exceeding one trillion, capable of processing images, videos, documents and other content forms within the same system.

He expects that the model will fully surpass the previous generation Qwen 3.7-Max in complex productivity scenarios such as coding, full-stack development, data analysis, and office automation.

Although the 2.4 trillion parameter scale is eye-catching, analysts caution investors and developers to remain prudent.

Analysts pointed out in interviews that the total parameter scale does not directly reflect the parameter count actually involved in computation for each token, nor is it possible to deduce the true inference cost. As of now, Alibaba has not released any benchmark results, nor has it disclosed the number of active parameters, authorization agreements, or a complete model card.

It is worth noting that early Qwen series models largely used the Mixture-of-Experts (MoE) architecture—in which only a small portion of weights are activated during inference, allowing actual computing power and cost to be kept far below the nominal parameter scale. If Qwen3.8 Max uses similar design, the gap between its 2.4 trillion "total parameters" and the actual inference expense may also be significant.

"Civil War" within Alibaba's Investment Landscape

Analysts believe that the backdrop of this release reflects profound changes in the competitive landscape of China's AI ecosystem.

Just days before the Qwen3.8 Max debut, Alibaba-invested Moonshot AI released Kimi K3. With a parameter scale reaching 2.8 trillion, this model stood out in horizontal comparisons with products from Anthropic and OpenAI, causing global tech stocks to fluctuate for a time and breaking the record for "largest open model."

The sudden emergence of Kimi K3 marks that China's large model labs no longer see open weights as an optional add-on, but as a core bargaining chip to win the developer ecosystem. Alibaba holds wide shares in leading domestic AI startups; the internal struggle within its ecosystem for the "largest open model" has effectively turned into a self-competition within Alibaba's investment territory.

Whether Qwen3.8 Max can deliver on the ambitions expressed in the preview version after full weight release remains to be tested by benchmarks and the market.

Risk DisclaimerThe market is risky; investment requires caution. This article does not constitute individual investment advice, nor does it take into account the specific investment objectives, financial situation, or needs of individual users. Users should consider whether any opinions, viewpoints, or conclusions in this article are suitable for their particular circumstances. Investment decisions based thereon are at your own risk.