Internal chart compares how many tokens models need to center a div

BLUECOW009 · x · 2026-07-29

An internal evaluation compares how many output tokens different models need to correctly center a <div> in HTML/CSS.

The chart shows lower is better and ranks models by median first-response tokens on 100 randomized prompts. The graph includes Claude Opus 5, GPT-5.6 Sol, Gemini 3.1 Pro, Grok 4.5, Claude Fable 5, Kimi K3, GPT-5.6 Terra, and GPT-5.6 Luna. By this metric, GPT-5.6 Luna is the most token-efficient, while Claude Opus 5 needs by far the most tokens.

Original post →

More from Models

Models channel →