Alibaba opened broad access this Monday, August 3, to its largest artificial intelligence model to date, Qwen3.8-Max, with 2.4 trillion parameters, according to an announcement reported by The Verge. The Chinese company contends that the system competes with the best models from Anthropic and OpenAI, as well as Moonshot AI's Kimi K3.
The model will be available to developers worldwide through Alibaba Cloud APIs and QwenWork, the agentic work platform that launched its public beta on the same day, according to South China Morning Post. Qwen3.8-Max handles a context window of up to one million tokens and accepts text, image, and video. What distinguishes it from other frontier releases is its openness: Alibaba announced that next week it will publish the model's weights, the values that determine how it processes information, marking a return to open source after a period of proprietary models. For developers and startups across Latin America, including Mexico, an open frontier model lowers the cost barrier to building on cutting-edge AI without depending on a single closed provider.
On Arena.AI, the community model comparison platform, Qwen3.8-Max ranks only behind Fable 5 and three models from Anthropic's Opus family; in frontend programming, only two Opus models and Kimi K3 (which Moonshot released with open weights last week) outperform it. According to Alibaba's own benchmarks, the model matches and in some cases surpasses Fable 5. The announcement comes one week after Kimi K3's release and as ByteDance and MiniMax presented new video generators on Friday. Alibaba shares in Hong Kong rose 7% on Monday, to HK$125.20, according to South China Morning Post.
The next milestone is the publication of the model's weights, expected in the coming days. That release will determine whether the openness commitment holds and whether a 2.4-trillion-parameter frontier model can deliver on its promise of putting cutting-edge AI within reach of more developers, including those in Mexico.
This article was written with artificial intelligence assistance from verified sources and reviewed by a human editor before publication.