Four-task model comparison: Images 2.5 wins image generation

Nakul @wtf_nakul7Originally posted 2026.09.13

Open original on X

I tested the biggest AI models across 4 different tasks. And honestly… the results surprised me. Coding: Gemini vs GPT-6 Astra → Winner: Astra Image generation: Nano Banana Pro VS ChatGPT Images 2.5 → Winner: GPT Video generation: Omini vs Chinese models → Winner: Chinese models - Seedance 2.5 is seriously good Website design: Gemini 3.8 Flash vs Claude → Winner: Claude - Gemini needed way more prompts to get close to the exact design. There is no single AI that wins everything anymore