On the Apex Math Reasoning benchmark, Qwen3.7-Max scored 44.5, eclipsing Claude Opus-4.6 Max's score of 34.5 and DeepSeek ...