VPCT
VPCT (Visual Pattern Completion Test) · tests visual reasoning and pattern recognition by having models complete visual sequences and transformations.
The Frontier
Best score over time · one chart, every benchmark
Full rankings
22 models tested · sorted by score
| # | Model | Score |
|---|---|---|
| 1 | 86.5 | |
| 2 | 76.0 | |
| 3 | 58.9 | |
| 4 | 49.0 | |
| 5 | 38.0 | |
| 6 | 36.3 | |
| 7 | 28.0 | |
| 8 | 19.6 | |
| 9 | 17.5 | |
| 10 | 10.3 | |
| 11 | 10.0 | |
| 12 | 10.0 | |
| 13 | 9.7 | |
| 14 | 8.5 | |
| 15 | 7.0 | |
| 16 | 7.0 | |
| 17 | 5.8 | |
| 18 | 5.5 | |
| 19 | 2.5 | |
| 20 | 1.0 | |
| 21 | 1.0 | |
| 22 | 1.0 |
Score distribution
Where models cluster
Correlated benchmarks
Pearson r · original research
Benchmarks that track with VPCT
Pearson correlation across models scored on both benchmarks. Closer to 1 = strongly predictive.
Frequently asked
About VPCT
What does VPCT measure?
VPCT (Visual Pattern Completion Test) · tests visual reasoning and pattern recognition by having models complete visual sequences and transformations. 22 AI models have been tested on it. Scores range from 1.0 to 86.5 out of 100.
Which model leads on VPCT?
Gemini 3 Pro from Google DeepMind leads VPCT with a score of 86.5. The median score across 22 tested models is 10.0.
Is VPCT saturated?
No · the top score is 86.5 out of 100 (87%). There is still meaningful room for improvement on VPCT.
Does VPCT predict performance on other benchmarks?
Yes · VPCT scores correlate 0.88 with FrontierMath-2025-02-28-Private across 19 shared models. Models that do well on VPCT tend to do well on FrontierMath-2025-02-28-Private.
How often is VPCT data refreshed?
BenchGecko pulls updates daily. New model scores on VPCT appear as soon as they are published by Epoch AI or the model provider.
- Category
- Multimodal
- Max score
- 100
- Models
- 22
- Updated
- 2025-12-17
More multimodal benchmarks
Same category · related evaluations