cd /news/artificial-intelligence/two-stealth-models-just-appeared-on-… · home › topics › artificial-intelligence › article
[ARTICLE · art-172489] src=contralabs.com ↗ pub= topic=artificial-intelligence verified=true sentiment=· neutral

Two stealth models just appeared on the leaderboard, both ahead of Fable 5.2

Two unnamed stealth models debuted on the Frontier Design Benchmark leaderboard from Contra Labs, with the top-ranked model, Gestalt, placing ahead of the human golden reference and Claude Fable 5.2. The benchmark, built on 33 real client design briefs, 68 professional designers, 4,311 blind votes and 35 rubric dimensions, ranks Gestalt first, human golden second and Claude Fable 5.2 third, followed by Claude Opus 5.5, GPT-6.1 Sol and Claude Sonnet 5.5. Contra Labs reports its design critique rubric scores correlate with designers' ratings at r = 0.94, adding explainable quality judgment on top of pairwise preferences.

recommendation

Pair blind pairwise votes with a rubric scored on 35 dimensions; Contra Labs found rubric scores correlated with designer ratings at r = 0.94, making design quality explainable beyond raw Elo.

more in playbook: Evals and testing for LLM apps
read1 min views3 publishedNov 19, 2026
Two stealth models just appeared on the leaderboard, both ahead of Fable 5.2
Image: source

Golden · Bad Move Bad Move website15 outputsClaude Fable 5.2Website

Exploring the frontier of AI design capabilities through real client briefs from Contra’s marketplace, golden designs and model comparisons.

Leaderboard #

How frontier models rank in blind head-to-head votes on real client briefs from Contra’s marketplace.

Frontier

Design Benchmark

Elo rating from blind head-to-head votes #

  1. Gestalt
  2. Human golden
  3. Claude Fable 5.2
  4. Claude Opus 5.5
  5. GPT-6.1 Sol
  6. Claude Sonnet 5.5
  7. Kimi K3.1 Pro
  8. GPT-6 Astra
  9. Qwen3.8 Max
  10. Kimi K3
  11. GLM-5.3 Prime
  12. Gemini 4 Carbon
  13. Muse Spark 1.3
  14. Muse Wonder
  15. Grok 4.7
  16. Gemini 4 Argon
  17. DeepSeek V4 Pro

contralabs.com/frontierdesign

Real design projects, judged by designers and scored by rubrics #

Frontier Design Benchmark is the first benchmark built on real world, high value, and complex design projects. It is judged by professional designers in the Contra Labs network, ensuring the quality of evaluation.

In the benchmark, we also built a design critique rubric and a list of AI slop patterns. The rubric scores are highly correlated with the designers’ ratings of the models (r = 0.94). These additional metrics provide explainable design quality judgement on top of pairwise preferences.

  • 33
  • design briefs
  • 68
  • professional designers
  • 4,311
  • blind votes
  • 35
  • rubric dimensions

Real client briefs, each with its golden: the design set beside the models’ outputs as the bar.

View all

Your vote ranks the models #

Two model outputs for the same brief, live and side by side, without the models’ names. Pick the better one, or call a tie.

Vote on outputs

── more in #artificial-intelligence 4 stories · sorted by recency
── more on @contra labs 3 stories trending now
sponsored brought to you by zahid.host 4,200+ EU-deployed projects
reading about agents? ship yours in a single git push.

Run your AI side-project on zahid.host

EU-based hosting, git-push deploys, automatic HTTPS, no cold starts. Free tier with a custom domain — perfect for shipping the agent you just read about.

$git push zahid main
→ Live at https://your-agent.zahid.host ✓
Get free account → Pricing
from €0/mo · no card required
LIVE [news/two-stealth-models-j…] indexed:0 read:1min 2026-11-19 · —