Chinese technology company Alibaba announced Qwen3.8-Max on Monday, calling it the largest and most capable model in its Qwen series to date. The new multimodal model has 2.4 trillion total parameters and supports a context window of up to 1 million tokens, Alibaba Cloud said in a statement. It can process visual information, as well as text, and is designed for coding, research, workplace applications, and tasks requiring prolonged autonomous operation. Qwen3.8-Max uses a sparse mixture-of-experts architecture and activates only 95 billion of its parameters for each operation, allowing it to reduce computing costs and response times compared with dense models of a similar scale. The model is slightly smaller by total parameter count than Moonshot AI's Kimi K3, which was released in July with 2.8 trillion total parameters and 104 billion activated parameters. Both models offer native visual capabilities and context windows of about 1 million tokens. Alibaba said Qwen3.8-Max ranked fifth on the Text Arena model leaderboard and second on the Vision Arena, while placing fourth on a coding-focused arena. In internal testing, the model independently carried out a software-engineering project over 16 days, developing a framework that used feedback, testing, and log analysis to improve itself. The company said the model can also analyze lengthy documents, television series, and livestreams, turning the content into searchable knowledge bases. Other potential applications include editing