On November 2nd, Alibaba Cloud's latest flagship model, Qwen3-Max, launched its "Deep Thinking" mode on its official website. This feature significantly improves the efficiency of handling complex tasks by enhancing inference chain analysis and multi-step problem decomposition capabilities. Alibaba Cloud launched Qwen3-Max on September 24th, which is the largest and most powerful language model the Tongyi team has developed to date, boasting over 1 trillion parameters and 36TB of pre-trained tokens.
The official version of Qwen3-Max-Instruct further enhances its code and agent capabilities, achieving industry-leading levels in comprehensive benchmark tests covering knowledge, reasoning, programming, instruction compliance, human preference alignment, agent tasks, and multilingual understanding. The Tongyi team previously stated that Qwen3-Max-Thinking, when combined with tools and increased computing resources during testing, has achieved 100% accuracy in challenging inference benchmarks such as AIME 25 and HMMT.
