The company introduced the new AI model Qwen3-Max-Thinking, which specializes in solving logical problems and demonstrates high performance in complex calculations. The model was trained on large volumes of data using modern reinforcement methods, which significantly enhanced its ability to analyze information and self-reflect.
Qwen3-Max-Thinking can adaptively use various tools during operation. It independently connects search, memory, and a “code interpreter” without the need for manual configuration. This allows the model to effectively solve multi-step tasks and find optimal solutions in complex situations.
According to test results, Qwen3-Max-Thinking exceeded the performance of Gemini 3 Pro in logical thinking. It scored 98.0 points in HMMT for February, indicating its strong mathematical abilities. In agent search, the model received 49.8 points for HLE, confirming its effectiveness in performing multi-step tasks.
Users can apply Qwen3-Max-Thinking for solving complex mathematical problems, analyzing large volumes of data, and automating agent scenarios. Thanks to the combination of deep knowledge and flexibility in using tools, this model will be useful for researchers, developers, and enthusiasts working with AI.
