同一提示词跑 Grok、Gemini、GPT Image 2.5 对比
作者用同一提示词测三款生图模型，认为 GPT Image 2.5 微观细节最强，真实度评分 9.1 排第一。
Ran the exact same prompt through 3 of the best current image models: • Grok Imagine 2 • Gemini 3 Pro Image • GPT Image 2.5 Interesting differences: Grok → most convincing actual smartphone photo. The candid composition, lighting and imperfections feel very natural. Gemini → best overall scene composition. The camera is farther from the mirror, giving the bathroom and perspective more realism. GPT → strongest micro-detail. Skin pores, hair strands, fabric texture and small imperfections are noticeably detailed, although the result feels slightly more polished. My ranking for realism: 🥇 GPT: 9.1/10 🥈 Gemini: 8.2/10 🥉 Grok: 7.5/10 Grok makes it feel real, Gemini makes the scene feel real, GPT makes the details feel real And this was from one identical prompt, no model-specific optimization the prompt I’m using 👇 http://ugcmaster.pixclipper.workers.dev