Frontier Design Benchmark - Design Critique: leaderboard

The rubric score of the Frontier Design Benchmark, where models build working websites from 33 real client briefs: each output's desktop view is graded on the Design Critique Rubric, 35 items drawn from design principles in ten dimensions (hierarchy, layout, typography, colour, polish, imagery, brand fit, usability, copy and accessibility), and the score is out of 100. It counts execution mistakes a designer can point to, such as a label too dark to read, not taste: the publisher reports that it tracks the designers' blind-vote Elo across models (r = 0.94) while rating the human reference designs below several models that lose to them in the votes.

Metric: Design Critique Rubric Score (0-100). Source: contralabs.com. Saturation forecast: Around February 2027. 16 models tracked. Updated 19 Nov 2026.

Top models

#ModelSettingScoreOverall rank
1Claude Fable 5.283.8#1
2Claude Opus 5.583.5#9
3Gestalt83.3
4GPT-6.1 Sol79.4#15
5Claude Sonnet 5.579.3#24
6GPT-6 AstraUltra78.7#10
7Kimi K3.1 Pro72.8#35
8Gemini 4 Carbon68.4#5
9Kimi K362.7#61
10Qwen 3.8 Max60.1#66
11Grok 4.759.1#51
12GLM 5.3 Prime59#81
13Muse Spark 1.358.7#48
14Muse Wonder57.5#34
15Gemini 4 Argon57.1#18

No result here: #3 Gemini 4 Neon, #4 GPT-6.2 Astra.

Interactive version: theaggregate.ai/benchmark?slug=frontier-design-benchmark-design-critique · How It Works · Data refreshed daily, snapshot 2026-11-20.