Alibaba Launches New Reasoning Model Qwen3-Max-Thinking, Performance Matches Top AI Systems

New Reasoning Model Released

Recently, Alibaba released its new reasoning model Qwen3-Max-Thinking, featuring over a trillion parameters and 36T Tokens of pre-training data.

Performance Matches Industry Leaders

In various tests including factual knowledge, complex reasoning, and instruction following, the model performs on par with top systems like GPT-5.2-Thinking, ClaudeOpus4.5, and Gemini3Pro.

• Over a trillion parameters
• 36T Tokens of pre-training data
• Performance rivals leading AI models