Kaggle Pelican Arena: leaderboard

Kaggle Benchmarks pairwise human-voted arena (kaggle/pelican-arena): models generate an SVG of a pelican riding a bicycle (passing if it renders), and human voters compare passing outputs pairwise to fit an Elo-style arena rating with 95% confidence intervals. Preserves Kaggle's reasoning-effort run configurations; higher is better.

Metric: Arena rating (human votes). Source: www.kaggle.com. Saturation forecast: Around January 2028. 27 models tracked. Updated 21 Nov 2026.

Top models

#ModelSettingScoreOverall rank
1GPT-6 AstraMax1807.59#10
2GPT-6.1 SolxHigh1793.1#15
3Claude Sonnet 5.5Max1732.38#24
4Claude Opus 5.5Max1728.41#9
5Gemini 4 ArgonHigh1682.49#19
6Claude Fable 5.1Max1673.44#20
7GPT-6 SolMax1665.11#34
8Gemini 4 CarbonHigh1657.74#7
9Gemini 4 HydrogenHigh1630.52#16
10GPT-5.6 SolMax1628.69#37
11Qwen 3.8 MaxxHigh1591.88#63
12GLM-5.3 Flash1580.45#133
13Gemini 3.8 FlashMedium1577.43#43
14Claude Opus 5xHigh1575.45#26
15GLM-5.3Max1567.34#72

No result here: #1 Claude Fable 5.2, #4 GPT-6.2 Astra.

Interactive version: theaggregate.ai/benchmark?slug=kaggle-pelican-arena · How It Works · Data refreshed daily, snapshot 2026-11-21.