Alibaba says latest AI model Qwen3-Max rivals best offerings from OpenAI, Google
Alibaba Cloud announced the Qwen3-Max model, with over 1 trillion parameters, just hours ahead of the annual Apsara Conference in Hangzhou
The artificial intelligence and cloud computing unit of Alibaba Group Holding has officially unveiled its largest-ever AI model, with performance benchmarks that rival those of leading American models, in the latest sign of the Chinese tech giant’s push to narrow the gap with US peers in this field.
On Wednesday, Alibaba Cloud announced the Qwen3-Max model, with over 1 trillion parameters, just hours ahead of its annual Apsara Conference in Hangzhou, capital of eastern Zhejiang province. Parent Alibaba owns the South China Morning Post.
The Qwen3-Max model has enhanced coding and agent capabilities while achieving top performance across a suite of benchmarks including domain knowledge, reasoning, coding, agent tasks and multilingual understanding, according to Alibaba Cloud.
It cited self-reported benchmark tests to show that Qwen3-Max matched or surpassed the best models from domestic and global competitors such as Anthropic’s Claude Opus 4, DeepSeek V3.1, xAI’s Grok 4 and GPT-5 Pro from OpenAI.
Qwen3-Max is not open-sourced, but is accessible through Alibaba’s cloud computing platform.