Frontier AI models benchmark comparison

Gemini 4 dropped today and Claude beat it at 3D, games and web design on day one. Then Meta’s new model, dead last at everything visual, made the only film that actually hits. Same 5 prompts, 4 frontier models: 3D castle → Opus 5.5 Playable game → Opus 5.5 Emotional short film → Muse Wonder Homepage design → GPT-6 Astra Writing → Gemini 4 All 4 prompts and the full scorecard in the comments ↓