ProductWebGen - Webpage Instruction Following: leaderboard
Product webpage generation (ProductWebGen, Liu et al., 2026): 500 requests in 13 product categories, each a source product image with layout, style, content and visual-consistency instructions, answered with renderable HTML and showcase images. The score is the percentage of the webpage instruction clauses the generated HTML satisfies, judged per clause by GLM-4.5; higher is better.
Metric: Webpage instruction following (%; share of the layout, style and content clauses the generated HTML satisfies, GLM-4.5 judge, 500 samples). Source: arxiv.org. Saturation forecast: Estimated already saturated. 8 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Grok 4 | 87 | #306 |
| 2 | Claude Sonnet 4 | 84 | #397 |
| 3 | Gemini 2.5 Flash | 81 | #452 |
| 4 | GPT-4o | 78 | #741 |
No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #5 Claude Opus 5.5.
Interactive version: theaggregate.ai/benchmark?slug=productwebgen-webpage-instruction-following · How It Works · Data refreshed daily, snapshot 2026-11-12.