Gestalt AI中文

Frontier Design Benchmark · #1

Gestalt AI

Gestalt is an AI design model built for real-world websites and application interfaces. In the Frontier Design Benchmark by Contra Labs and BenchFlow, it ranked #1 in double-blind evaluations by professional designers.

Elo rating
1,931
#1 on leaderboard
Blind win rate
90%
95% CI: 86–92%
Elo vs humans
+192
+218 vs Claude Fable 5.2
Top 8 leaderboard snapshot
RankEntryEloGapWin rate95% CI
#1Gestalt (stealth)1,931Top90%86–92%
#2Human designers1,739-19274%71–78%
#3Claude Fable 5.21,713-21873%69–76%
#4Claude Opus 5.51,691-24073%—
#5GPT-6.1 Sol1,668-26370%66–74%
#6Claude Sonnet 5.51,667-26469%65–73%
#7Kimi K3.1 Pro1,663-26871%66–74%
#8GPT-6 Astra1,638-29366%61–70%

The November 18, 2026 research report recorded Gestalt at 1,928 Elo. The table above reflects the Top 8 live leaderboard as checked on November 21, 2026. Win rate reflects overall blind head-to-head comparisons. Research report

Why Gestalt AI stands out in design

90% win rate · 1,931 Elo

Leading blind preference

Across 4,311 blind head-to-head votes, Gestalt outperformed human reference designs (+192 Elo) and leading commercial models including Claude Fable 5.2 (+218 Elo) and GPT-6.1 Sol (+263 Elo).

Score 83 · r = 0.94

35-dimension Design Critique Rubric

On 35 criteria covering structure, visual craft, content, and usability, Gestalt (83) matches Claude Fable 5.2 and Opus 5.5 (84) and exceeds the human baseline (74), while leading strongly in whole-page taste.

Slop Detector · 0–100

Low AI slop signature

The benchmark’s Slop Detector measures generic AI layout tells such as repetitive card grids, predictable typography, and awkward spacing. Gestalt sits close to human reference designs at the low end.

Methodology

Rather than one-line toy prompts, the Frontier Design Benchmark tests 33 commercial briefs—27 adapted from real Contra client commissions and 6 from startup partners—judged double-blind by 35 professional designers (median 8+ years of experience).

  • Marketing websites
  • Landing pages
  • Web applications
  • Mobile interfaces
Real design briefs
33
Pro designers (8+ yrs)
35
Blind votes cast
4,311

Frequently asked questions

What types of design tasks does Gestalt AI excel at?

In the Frontier Design Benchmark, Gestalt is evaluated on full commercial briefs across four categories: marketing websites, landing pages, web applications, and mobile interfaces, each with brand assets and multi-section requirements.

How large is Gestalt’s lead over human designs, Claude Fable 5.2, and GPT-6.1 Sol?

On the November 21, 2026 leaderboard snapshot, Gestalt leads at 1,931 Elo (90% win rate, 95% CI 86–92%), ahead of human reference designs (1,739 Elo, 74%), Claude Fable 5.2 (1,713 Elo, 73%), Claude Opus 5.5 (1,691 Elo, 73%), and GPT-6.1 Sol (1,668 Elo, 70%).

What do the 35-dimension Rubric and Slop Detector measure?

The Design Critique Rubric checks 35 concrete criteria across structure, visual craft, content, and usability (correlating at r = 0.94 with expert preference). The Slop Detector scores generic AI layout clichés on a 0–100 scale, where Gestalt scores near human baselines.

Where can I inspect Gestalt’s generated websites?

Contra Labs publishes side-by-side benchmark outputs in its official Design Gallery (contralabs.com/frontierdesign/gallery). Gestalt is currently tested in stealth with no public API yet.

Primary sources