On January 26, 2026, Alibaba unveiled its premier reasoning model, Qwen3-Max-Thinking. This model is characterized by an impressive total parameter count that surpasses one trillion, coupled with a pre-training dataset comprising 36T Tokens. It has witnessed substantial enhancements in crucial areas, including factual knowledge, intricate reasoning, instruction adherence, alignment with human preferences, and agent functionalities. Across 19 authoritative benchmark tests, its performance stands on par with leading models such as GPT-5.2-Thinking, Claude-Opus-4.5, and Gemini 3 Pro. Qwen3-Max-Thinking brings forth two pivotal innovations: an adaptive tool-calling feature that facilitates the on-demand utilization of search engines and code interpreters, now integrated into Qwen Chat; and test-time scaling techniques that notably boost reasoning prowess.
