EuroEval Polish Summarization - PSC: leaderboard
EuroEval's Polish summarization task on the PSC dataset — models write abstractive summaries of Polish news articles, scored with automatic text-overlap metrics, from the EuroEval European language model benchmark suite.
Metric: chrF3++ (%). Source: euroeval.com. Saturation forecast: Not forecast. 148 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Ministral-3-3B-Instruct-2512 | 25.82 | #855 |
| 2 | NVIDIA-Nemotron-3-Nano-30B-A3B-BF16 | 25.68 | #679 |
| 3 | Ministral-3-8B-Instruct-2512 | 25.67 | #661 |
| 4 | GPT-OSS-20B (Low) | 25.36 | #518 (GPT-OSS-20B) |
| 5 | GPT-OSS-20B (Medium) | 25.35 | #518 (GPT-OSS-20B) |
| 6 | GPT-5.4 Nano (2026-03-17) (High) | 25.24 | #266 (GPT-5.4 Nano (2026-03-17)) |
| 7 | Phi-4-reasoning-plus (Thinking) | 25.2 | |
| 8 | LFM2 1.2B | 25.18 | #1503 |
| 9 | Apertus-70B-Instruct-2509 | 25.01 | #820 |
| 10 | EuroLLM-9B-Instruct | 25.01 | #1054 |
| 11 | LFM2-700M | 24.94 | #1555 |
| 12 | Ministral-3-14B-Instruct-2512 | 24.73 | #615 |
| 13 | Bielik-11B-v2.3-Instruct | 24.62 | #693 |
| 14 | Claude Haiku 4.5 (20251001) | 24.61 | #261 |
| 15 | Qwen 3 1.7B (Non-reasoning) | 24.52 | #1196 (Qwen 3 1.7B) |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=euroeval-polish-summarization-psc · How It Works · Data refreshed daily, snapshot 2026-10-20.