Alibaba’s Qwen team has previewed Qwen3.8-Max-Preview, its next flagship AI model, as China’s frontier AI race continues to accelerate.
The company describes the model as a 2.4 trillion-parameter system and says it is second only to Anthropic’s Fable 5 among the models it tested. However, Alibaba has not yet published the full benchmark table, model card, license, or open-weight release date, so its performance claims remain company-provided for now.
Preview Is Live
The preview version of Qwen3.8-Max is already available for testing.
Alibaba said the model is live on its coding platforms, while reports say access is available through Alibaba’s Token Plan, Qoder, and QoderWork at a preview discount. The company also says it plans to release the model as open-weight soon, but it has not given a date or shared a license file yet.
The announcement came during the World AI Conference in Shanghai and arrived shortly after Moonshot AI released Kimi K3, a 2.8 trillion-parameter open-weight model. That timing has made the launch part of a wider battle between Chinese AI labs for open-weight model leadership.
First Trillion-Parameter Multimodal Qwen Model
Qwen developer Shuai Bai described Qwen3.8-Max-Preview as the team’s first trillion-parameter multimodal model.
The model is said to handle text, images, video, and documents. Alibaba’s team also claims it improves on Qwen3.7-Max in coding, full-stack development, data analysis, and office workflows. However, independent tests have not yet confirmed those claims.
The 2.4 Trillion-Parameter Claim Needs Context
The headline number is large, but it does not fully explain the model’s real-world cost or speed.
Qwen’s earlier open models include sparse Mixture-of-Experts designs. For example, Qwen3-235B-A22B has 235 billion total parameters but activates 22 billion per token, while Qwen3-30B-A3B activates around 3 billion per token. That difference matters because active parameters, not just total parameters, strongly affect inference cost.
Alibaba has not disclosed the active-parameter count for Qwen3.8-Max-Preview. Without that figure, the 2.4 trillion-parameter claim says little about how expensive the model will be to run.
Developers Are Waiting for the Real Release
The reaction from developers has been mixed.
Many users welcomed another large Chinese open-weight model, especially after the recent Kimi K3 launch. Others pointed out that a live preview is not the same as an open-weight release. For developers, the important details are still missing: the model card, benchmark methodology, license, repository, active-parameter count, and final API pricing.
Alibaba’s current public pricing page lists existing Qwen Max and Qwen3.7 Max pricing, but it does not yet provide a complete standard pricing page for Qwen3.8-Max-Preview.
Production Users Should Wait
Qwen3.8-Max-Preview may become one of Alibaba’s most important AI releases, but businesses should be careful before moving production workloads.
The preview is available, but the lack of a published benchmark table, model card, license, and open-weight repository means developers still need to test it on their own workloads before making decisions.
For now, Alibaba has delivered a major AI teaser: a 2.4 trillion-parameter multimodal model that it says can compete near the top of the frontier model market. The real test will come when independent benchmarks, pricing, active-parameter details, and downloadable weights become available.
Stay Connected with ProPakistani
Get the latest tech news, telecom insights, and product launches wherever you prefer.
Add ProPakistani to Preferred Sources and see more of our stories in Google Search and Top Stories.
