Polish Cultural Vision Benchmark - Object: leaderboard
SpeakLeash's Polish cultural imagery suite: given a photograph of something culturally Polish — a dish, a building, a monument, a tradition — name the object. Spans art and entertainment, history, culture and tradition, and geography. The harder of the benchmark's two questions by roughly 20 points on every model.
Metric: Accuracy (%). Source: huggingface.co. Saturation forecast: Around September 2027. 42 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemini 3 Flash | 42.24 | #99 |
| 2 | Gemini 3.1 Pro (Preview) | 39.45 | #58 |
| 3 | Gemini 3.1 Flash Lite | 33.08 | #191 |
| 4 | Claude Opus 4.6 | 32.65 | #63 |
| 5 | Gemini 2.5 Pro | 31.36 | #152 |
| 6 | Gemini 2.5 Flash | 28.78 | #245 |
| 7 | Seed 2.0 Mini | 27.8 | #276 |
| 8 | Claude Sonnet 4.6 | 27.39 | #90 |
| 9 | Claude 3.7 Sonnet | 22.39 | #249 |
| 10 | Gemini 2.5 Flash Lite | 21.46 | #422 |
| 11 | GPT-5.4 | 21.16 | #79 |
| 12 | GPT-5 Mini | 17.71 | #186 |
| 13 | Claude Sonnet 4.5 | 17.51 | #142 |
| 14 | GPT-4o | 17.35 | #343 |
| 15 | Qwen 3.5 Plus (02-15) | 16.11 | #218 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=polish-cultural-vision-benchmark-object · How It Works · Data refreshed daily, snapshot 2026-10-20.