[tmaly]

Over the weekend I was running the same prompt across GPT-5.2, Gemini 3, and Grok. Both Gemini 3 and Grok on thinking mode finished within 2 minutes. GPT-5.2 was just spinning its wheels for like 6 minutes.