Sonnet 5.5 and GPT-6.1 Sol take the same candle-in-the-wind animation prompt
Given the same prompt, Claude Sonnet 5.5 built a full candlelit room where a storm blows the candle out, while GPT-6.1 Sol made a minimal lab scene with a flame that would not go out; the post compares tokens, time and frame rate.
Claude Sonnet 5.5 vs GPT-6.1 Sol (both medium) Same prompt: "a realistic candle that reacts to your cursor like wind." One shot each. Then I sent a storm at both. 🕯️ Sonnet built a whole candlelit room. Shadows of the bottle and books swing across the wall as the flame bends, and the storm blew it out. 🕯️ GPT built a sleek minimal lab study. The flame bent to 9% and refused to die. Sonnet: 194k tokens, 15 min, 65 fps GPT: 58k tokens, 8 min, 38 fps Both used 9% of their usage limit. Which candle would you light? 👇