Alibaba released a preview of its flagship Qwen3.8 Max on Jul. 19, a 2.4 trillion-parameter model the company claims trails only Anthropic's Fable 5. The Hangzhou-based company posted the announcement, calling Qwen3.8 Max one of the most powerful models available today. Developers can test the preview through Alibaba's coding platforms, including Qoder, while a live chat version is open to the public. Full open weights will follow soon, the company said, without naming a date. The Qwen team expects the model to outperform its predecessor, Qwen 3.7-Max, in coding and complex productivity work such as full-stack development, data analysis and office workflows, developer Shuai Bai said. Bai called it the team's first multimodal model above 1 trillion parameters, one capable of processing images, videos and documents in a single system. No benchmark results have been published so far, and the launch post offered no technical specifications beyond the parameter count. Analysts cautioned that the headline figure deserves scrutiny, since total parameter count says little about how much compute a model actually uses per token, or what it costs to run. Alibaba has not disclosed an activated parameter count, license terms or a model card. The preview arrived days after Moonshot AI, a startup backed by Alibaba, released Kimi K3, a model whose strong showing against systems from Anthropic and OpenAI shook global technology stocks. Kimi