Opus 5.5 vs GPT-6 Astra in 3D: a week of testing shows a clear split
A week of testing found Opus 5.5 better at 3D assets, animation and game UI but slower and more token-hungry, while GPT-6 Astra finished 20-45 minutes earlier, used about 10x fewer thinking tokens and produced more organized scenes.
Claude Opus 5.5 vs GPT-6 Astra in 3D... what each one is actually good at after this week's tests (and mine, video below), the split is clear: Opus 5.5: - better 3D assets: shapes, textures, lighting - more alive animation, more range in motion - stronger UI and sound when building 3D games - thinks longer, burns more tokens, takes more time GPT-6 Astra: - faster, finished 20-45 min earlier in most tests - ~10x fewer thinking tokens - more organized scenes: lights, controls, cameras set up like a real production file - cleaner, more "designed" look, but plainer assets so: > hero asset, character, game prototype -> Opus > fast iteration, clean scene setup, tight budget -> Astra and one catch: Opus is 60% cheaper per token but can still cost more per build, because it thinks so much longer
This post quotes another post: x.com/0xmovez/status/2104216919033192746