# Three.js Resources — Full AI Hub Content (llms-full.txt)

## Benchmark Methodology
Each test uses a single fixed prompt sent verbatim to each model (one-shot, no follow-up). Prompt version is frozen; any change bumps the version and invalidates old comparisons. Five criteria scored 0–10: runs-first-try, visual quality, code quality, performance, prompt adherence. Max 50 per test.

## Benchmark Prompt Texts