Alibaba teases new Qwen previews, highest-ranking Chinese AI models on Arena

Alibaba Group Holding has released preview versions of its upcoming Qwen3.7 AI models, marking a continued push into the competitive AI sector. The Qwen3.7-Max-Preview and Qwen3.7-Plus-Preview ranked 13th and 16th globally in text and vision capabilities, respectively, according to benchmarking platform LM Arena. These rankings position Alibaba’s models behind industry leaders like Anthropic’s Claude, Google’s Gemini, and OpenAI’s ChatGPT. The previews follow the launch of Alibaba’s Qwen3.6 series last month and precede a major announcement at the Alibaba Cloud Summit in Hangzhou on Wednesday. The company has adopted a strategy of releasing preview versions on Arena to gather user feedback and refine performance before final releases. This approach aligns with industry trends where companies like Alibaba are moving away from fully open-sourcing their most advanced models. Alibaba’s shift toward proprietary models reflects broader industry pressures, including the high costs of computing power required for AI development. The Qwen3.6-Max and Qwen3.6-Plus models were already proprietary, accessible only through paid API access. This commercialization trend is driven by the need to monetize AI advancements while maintaining competitive performance. The Qwen3.7 previews highlight Alibaba’s efforts to strengthen its AI capabilities, particularly in text and vision tasks. While the models are not yet finalized, their current rankings suggest progress in narrowing the gap with leading global AI systems. The company’s focus on proprietary models also signals a strategic pivot toward sustainable revenue generation in the AI market.