Worlds by model

Claude Opus 5.5. Every 3D world it built.

Claude Opus 5.5 by Anthropic, run in Claude Code: every AI-generated picture it built as a Three.js world, and every fix experiment it took part in. One run per build, with no human edits.

Made by
Anthropic
Agent
Claude Code
Island cabin builds
low, medium, high and extra high
Fix experiments
three
Other pictures
one

These are not CmdBench benchmark runs: no checks, scores or verdicts.

AI-generated picture of a miniature island: a wooden cabin with a stone chimney, a jetty and rowboat, a yellow crane over a half-built timber frame, lumber stacks, rocks and pine trees, in warm late-afternoon light.
Reference picture, AI-generated. Every island cabin build got this picture and the same prompt.

The island cabin picture

Claude Opus 5.5 built the same AI-generated picture as every model in the grid, from the same prompt: one run on each effort level, low, medium, high and extra high. Each picture opens that world full screen.

Costs are Claude Code's own estimates at API list prices.

Fix experiments

A model builds the island cabin picture, then a fresh session improves a copy of that build from one fix prompt and one screenshot of the build's opening view. Claude Opus 5.5 built or fixed builds in these three experiments.

Other pictures

The same test on a different AI-generated picture, with the same kind of prompt.