Bongard in Wonderland — leaderboard
Visual abstract reasoning benchmark using classic Bongard problems. Models must identify discriminative visual concepts from sets of positive and negative examples. Hardcoded from arXiv:2410.19546.
Metric: Problems Solved (out of 100). Source: ml-research.github.io. Status: saturation imminent. 10 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | GPT-5 | 64 |
| 2 | O3 | 53 |
| 3 | O1 | 43 |
| 4 | Claude 3.5 Sonnet | 31 |
| 5 | GPT-4o | 25 |
| 6 | Gemini 2.0 Flash | 25 |
| 7 | Gemini 1.5 Pro | 8 |
Interactive version: theaggregate.ai/benchmark?slug=bongard-in-wonderland · How the rankings work · Data refreshed daily, snapshot 2026-07-22.