Polish Knowledge Benchmark - Media & Journalism: leaderboard
Topic category of the Polish Knowledge Benchmark (PolishQA) covering Polish media and journalism. The benchmark has 15,000 synthetic four-option questions about Poland generated by DeepSeek 3.2 and difficulty-rated by Qwen3.5-397B-A17B. Accuracy is correct over total in the category, 0-100, higher is better.
Metric: Accuracy (%). Source: huggingface.co. Saturation forecast: Around February 2027. 69 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemini 3 Flash (Preview) | 90.26 | #90 |
| 2 | Gemini 3.6 Flash | 90 | #57 |
| 3 | Gemini 3.5 Flash Lite | 88.68 | #174 |
| 4 | Gemini 3.1 Flash Lite (Preview) | 87.37 | #201 |
| 5 | Gemini 3.1 Flash Lite | 86.84 | #200 |
| 6 | Mistral Large 4 | 81.58 | #106 |
| 7 | DeepSeek V4 Flash | 81.32 | #155 |
| 8 | Gemini 2.5 Flash | 80.26 | #255 |
| 9 | Gemini 2.0 Flash (001) | 79.47 | #365 |
| 10 | Mistral Large 3 | 78.95 | #410 |
| 11 | Qwen 3.5 397B A17B | 78.42 | #154 |
| 12 | DeepSeek V3.2 | 78.16 | #210 |
| 13 | Mistral Medium 3.1 | 78.16 | #418 |
| 14 | Mistral Small 4 | 77.37 | #522 |
| 15 | Grok 4.1 Fast | 77.11 | #227 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=polish-knowledge-benchmark-media-journalism · How It Works · Data refreshed daily, snapshot 2026-11-05.