GPT-4.5 (Preview)
Provider: OpenAI • ID: gpt-4-5
Detailed evaluation metrics and judge breakdown for GPT-4.5 (Preview) across all 4 launch categories.
Compare in Blind Arena
Pricing: In — • Out —
Composite Verdict Score/100
13.6
0
.0
Category Score Breakdown
Frontend UI13.9
Interactive dashboards and landing pages
Game Dev13.3
Canvas 2D physics and browser games
SVG Art13.7
Vector graphics and generative math art
Agentic Tasks13.2
Multi-step execution planning
Auditable Multi-Judge Dimension Scoring
Grades evaluated across 3 independent judge models using published rubrics.
Functionality— Executes without runtime JS exceptions
14.0
Craft— Clean semantic markup and clean TypeScript
13.7
Design— Visual aesthetics, contrast, layout harmony
13.5
Creativity— Original micro-interactions and layout depth
13.2
Fidelity— Strict adherence to complex prompt constraints
13.3