Landing bench 3B: AI-generated imagery

Blind critic: Grok 4.7 (critic cost $0.126). Contact sheet

Screenshot of claude-haiku-5.5 page1. claude-haiku-5.56.9/10 · gen $0.0083 + img $0.3362 · 1 attMosaic metaphor actually drives prompts, steps, and studies.Screenshot of glm-5.3-flash page2. glm-5.3-flash6.8/10 · gen $0.004 + img $0.3362 · 1 attLoad-bearing concept locked to palette, lattice, and lab prompts.Screenshot of muse-spark-1.3 page3. muse-spark-1.36.8/10 · gen $0.0324 + img $0.3362 · 1 attWork-order conceit and chaos dial that really restyles the page.Screenshot of qwen3.8-flash page4. qwen3.8-flash6.6/10 · gen $0.0126 + img $0.3362 · 1 attTest-rig voice with careful reduced-motion and pointer tilt.Screenshot of gpt-6-luna page5. gpt-6-luna6.1/10 · gen $0.0032 + img $0.2018 · 1 attQuiet type, crop, and contrast discipline.Screenshot of mimo-v2.6-flash page6. mimo-v2.6-flash5.5/10 · gen $0.013 + img $0.4035 · 4 att · fixLoud color system and confident display rhythm.