Creative Showcases

Standardized creative HTML generation tasks. Each model receives the exact same prompt and the output is evaluated by a judge model on visual quality, interactivity, code completeness, and creativity. These results are separate from the main public-benchmark leaderboard scores.

4 Completed results 3 Models with results

Creative Scoreboard

Average across completed creative showcases
# Model Creative Avg Showcases
1 OpenAI: gpt-oss-120b OpenAI 7.8 /10 1
2 Llama 3.3 70B Meta 7.7 /10 2
3 Llama 4 Scout Meta 3.6 /10 1

3D Safari Simulation

Creative-html

Generate an interactive 3D safari experience in a single HTML file using Three.js or pure CSS 3D.

OpenAI: gpt-oss-120b
OpenAI
7.8/10
Visual Quality 7.5/10 Interactivity 7.5/10 Completeness 9.0/10 Creativity 7.0/10 View full output →
Llama 3.3 70B
Meta
7.0/10
Visual Quality 7.0/10 Interactivity 7.0/10 Completeness 7.0/10 Creativity 7.0/10 View full output →

Web3 Landing Page

Creative-html

Generate a complete, visually striking single-file HTML landing page for a fictional Web3 crypto project.

Llama 3.3 70B
Meta
8.4/10
Visual Quality 9.0/10 Interactivity 8.5/10 Completeness 9.0/10 Creativity 6.5/10 View full output →
Llama 4 Scout
Meta
3.6/10
Visual Quality 4.0/10 Interactivity 3.0/10 Completeness 4.0/10 Creativity 3.0/10 View full output →
Learn about the overall scoring methodology →