Opus 5.5 beats GPT-6 Sol on feature depth in a one-file game test
Given the same one-prompt, one-file game challenge, Sol was faster and cleaner but produced one fish and two upgrades, while Opus built five fish species, rarity tiers and a save system.
Opus 5.5 just crushed GPT-6 Sol in my latest game-dev test — and it wasn’t close. Both got the same challenge: One prompt. One HTML file. Full autopilot. GPT-6 Sol was faster and produced a smaller, cleaner build. But Opus built the better game. 🐟 Sol • 1 generic fish • 2 basic upgrades • Simple gameplay loop • Polished tech-demo feel 🎣 Opus • 5 distinct fish species • Rarity tiers • Perfect-catch mechanics • 3 progression trees • Dynamic fish schools • Minimap • Persistent save system That’s the difference that stood out to me: Sol optimized the benchmark. Opus understood the product. FPS and file size look great on a chart. But when the goal is to actually build a game people want to play, feature depth and systems matter a lot more. Full side-by-side below 👇