ع
Alibaba unveiled its largest AI model yet on Monday, the Gulf Times reported, sending the company’s shares up 7 percent in Hong Kong trading as the new system immediately climbed AI leaderboards. The Qwen3.8-Max features 2.4 trillion parameters and employs a mixture-of-experts architecture that activates only 95 billion parameters per task to control costs and speed. Alibaba stated the model completed a software engineering project in 16 days and can process up to 1 million tokens including text, images and video. The tech giant plans to release the open weights next week, marking the first time it has open-sourced a Qwen-Max class model.
According to Arena.AI leaderboard data, Qwen3.8-Max ranks as the top Chinese model for text tasks while placing fifth overall behind several Anthropic variants including Claude Fable 5. The same platform’s vision leaderboard positions the Alibaba system second globally, trailing only one Claude Fable 5 variant. Alibaba’s announcement follows its February 2026 release of the Qwen3.5 series that introduced agentic capabilities, a Reuters report noted. The new model joins Moonshot AI’s Kimi K3, which features 2.8 trillion parameters, in the upper tier of Chinese systems measured by scale.
DeepSeek released its V4-Flash model on Friday, the Gulf Times stated, achieving the lowest inference cost among major systems tracked by Artificial Analysis. The startup charges 14 cents per million input tokens and 28 cents per million output tokens, resulting in an average test cost of 3 cents compared with 86 cents for Moonshot’s Kimi K3 and $3.15 for Anthropic’s Claude Fable 5. Artificial Analysis figures show the DeepSeek offering runs more than 100 times cheaper than certain premium models on benchmark tasks when accounting for total processing steps. The company previously gained attention in early 2025 when its R1 and V3 models triggered a global tech stock selloff.
Both the Alibaba and DeepSeek releases emphasize open-weight designs that make underlying parameters available for download and adaptation, the Gulf Times reported. This approach contrasts with closed-source systems from OpenAI, Anthropic and Google. Chinese developers have increasingly targeted workflows that require capable but not frontier-level performance at lower prices. The models process large contexts such as legal documents or codebases efficiently through their expanded token windows.
Omdia chief analyst Lian Jye Su said Chinese AI companies have found an important market. “Many business workflows do not need the industry’s very best model,” Su added. “They need models that are good enough, affordable, transparent and accessible, and open-weight models help meet that demand.” The analyst’s assessment aligns with broader industry shifts toward cost-effective systems that prioritize accessibility for global developers. DeepSeek is preparing for a potential initial public offering, sources told the Gulf Times.
Alibaba’s share performance followed a similar 5.4 percent rise in July when the company first previewed the Qwen3.8-Max, a Bloomberg report indicated. The latest model builds on the Qwen series that has progressively added multimodal capabilities across text, audio and vision. Industry tracking shows Chinese firms continue to advance parameter counts and efficiency metrics while maintaining open-weight strategies to capture developer adoption outside premium closed models.
