EuroEval Serbian Summarization - LR SUM SR: leaderboard
EuroEval's Serbian summarization task on the LR SUM SR dataset — models write abstractive summaries of Serbian 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 | Bielik-11B-v2.3-Instruct | 31.95 | #693 |
| 2 | Llama 3.1 8B Instruct | 31.53 | #1035 |
| 3 | Mistral Small 3.1 | 31.09 | #610 |
| 4 | Llama 3.3 70B Instruct | 31.08 | #534 |
| 5 | occiglot-7B-eu5-instruct | 30.87 | |
| 6 | Qwen 3 14B (Non-reasoning) | 30.64 | #537 (Qwen 3 14B) |
| 7 | Gemma 3 270M (IT) | 30.6 | #1581 |
| 8 | Voxtral-Small-24B-2507 | 30.49 | #581 |
| 9 | Mistral Small 3.2 | 30.38 | #605 |
| 10 | Qwen 3 4B (Non-reasoning) | 30.35 | #829 (Qwen 3 4B) |
| 11 | Gemma 3 12B (IT) | 30.34 | #664 |
| 12 | Apertus-70B-Instruct-2509 | 30.32 | #820 |
| 13 | Qwen 3 30B A3B 2507 Instruct | 30.3 | #488 |
| 14 | tiny-aya-water | 30.28 | |
| 15 | Qwen 2.5 72B Instruct | 30.27 | #454 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=euroeval-serbian-summarization-lr-sum-sr · How It Works · Data refreshed daily, snapshot 2026-10-20.