CounterCount: leaderboard
Vision-Language Models (VLMs) excel at multimodal reasoning, yet it remains unclear whether their answers are grounded in visual evidence or driven by learned language and world priors. Counting provides a precise testbed: when visual evidence conflicts with canonical object knowledge, a model must rely on the image rather than a prototypical count. We introduce CounterCount, a diagnostic framewor
Metric: Avg Acc (self-reported). Source: benchmarklist.com. 7 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemma 3 4B | 44.57 | #1089 |
| 2 | Gemma 3 12B | 40.22 | #671 |
| 3 | Gemma 3 27B | 36.96 | #607 |
| 4 | Claude Haiku 4.5 | 36.06 | #280 |
Interactive version: theaggregate.ai/benchmark?slug=countercount · How It Works · Data refreshed daily, snapshot 2026-10-20.