Alibaba AI Qwen latest model beats Google Gemini and ChatGPT in coding

Alibaba has introduced Qwen3.7-Max, an AI model that outperforms Google’s Gemini and OpenAI’s ChatGPT in coding benchmarks, achieving fourth place globally on Code Arena with a score of 1,541. Unlike traditional chatbots, Qwen3.7-Max is built for autonomous, agent-based workflows, capable of handling long coding tasks independently for up to 35 hours. The model demonstrated its capabilities by optimizing code for Alibaba’s AI chips, running 432 kernel tests and making over 1,100 tool calls to achieve a 10x performance improvement without prior exposure to the chip architecture. The model supports OpenAI- and Anthropic-compatible interfaces and integrates with tools like Claude Code, OpenClaw, and Qwen Code. Alibaba emphasizes its versatility beyond coding, including monitoring AI training systems, detecting suspicious behavior during software tests, and guiding robots through physical spaces. While earlier Qwen models were open-source, Qwen3.7-Max is proprietary and accessible exclusively through Alibaba Cloud’s Model Studio API. Alibaba claims Qwen3.7-Max performed strongly across reasoning and coding benchmarks, nearing Anthropic’s Claude Opus 4.6 Max in several tests. The company notes that benchmark results are self-reported and plans to release a detailed technical report later. This launch reflects China’s push to compete with U.S. AI leaders in autonomous coding agents, positioning Qwen3.7-Max as a key player in advanced AI development.