BabyVision: leaderboard
Visual reasoning tasks that 3-year-old children can solve - fine-grained discrimination, visual tracking, spatial perception, and pattern recognition. 388 questions across 22 subtypes. Human baseline: 94.1%.
Metric: Score (%). Source: unipat.ai. Source: llm-stats.com. Saturation forecast: Around January 2027. 20 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Median Human | 94.1 | |
| 2 | DeepSeek V4.1 Flash (1017) | 92.5 | #20 |
| 3 | MiniMax-M3.1-Flash | 91 | #33 |
| 4 | DeepSeek V4.1 Flash | 89.6 | #36 |
| 5 | Kimi K3 | 85.7 | #30 |
| 6 | Qwen 3.8 27B | 85.6 | #105 |
| 7 | Qwen 3.8 Max | 82 | #35 |
| 8 | Muse Spark Open | 80.2 | #54 |
| 9 | Muse Spark 1.1 | 76.3 | #52 |
| 10 | Seed 2.1 Pro | 73.7 | #67 |
| 11 | Qwen 3.7 Plus | 70.4 | #92 |
| 12 | Kimi K2.6 | 68.5 | #104 |
| 13 | Seed 2.1 Turbo | 62.9 | #86 |
| 14 | Seed 2.0 Pro (High) | 60.6 | #118 (Seed 2.0 Pro) |
| 15 | GLM-5.3 Flash | 53.4 | #65 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=babyvision · How It Works · Data refreshed daily, snapshot 2026-10-20.