Gemini 3 Pro and GPT 5.2 are amazing models but they are not the only games in town. We now have new Kimi and Qwen models that are on par in performance with those in many areas. Take the Qwen3-Max-Thinking for instance: it is trained with advanced RL to deliver great performance with reasoning, tool use, and agents. With multi-round self-reflection, it can beat top models on reasoning.
Qwen3-Max-Thinking automatically selects and leverages its built-in Search, Memory, and Code Interpreter capabilities as you chat with it to provide results without hallucinations. You can now access this through the chat interface but the API is also available.
🚀 Introducing Qwen3-Max-Thinking, our most capable reasoning model yet. Trained with massive scale and advanced RL, it delivers strong performance across reasoning, knowledge, tool use, and agent capabilities.
✨ Key innovations:
✅ Adaptive tool-use: intelligently leverages… pic.twitter.com/6sZiKWQAq3— Qwen (@Alibaba_Qwen) January 26, 2026
[HT]
