EuroEval Hungarian Summarization - Hunsum: leaderboard
EuroEval's Hungarian summarization task on the Hunsum dataset — models write abstractive summaries of Hungarian news articles, scored with automatic text-overlap metrics, from the EuroEval European language model benchmark suite.
Metric: chrF3++ (%). Source: euroeval.com. Saturation forecast: Around 2035. 149 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-4.1 | 34.8 | #248 |
| 2 | Gemini 3 Pro (Preview) | 34.53 | #66 |
| 3 | Apertus-70B-Instruct-2509 | 33.96 | #820 |
| 4 | Gemini 3 Flash (Preview) (Non-reasoning) | 33.91 | #82 (Gemini 3 Flash (Preview)) |
| 5 | Mistral Small 3.1 | 33.66 | #610 |
| 6 | Gemini 2.5 Flash Lite (Non-reasoning) | 33.65 | #422 (Gemini 2.5 Flash Lite) |
| 7 | Gemma 3 27B (IT) | 33.49 | #523 |
| 8 | Mistral Small 3.2 | 33.46 | #605 |
| 9 | Gemma 4 31B (IT) | 33.42 | #176 |
| 10 | EuroLLM-9B-Instruct | 33.42 | #1054 |
| 11 | GPT-5.4 Mini (2026-03-17) (High) | 33.39 | #169 (GPT-5.4 Mini (2026-03-17)) |
| 12 | Claude Sonnet 4.5 (Non-reasoning) | 33.37 | #142 (Claude Sonnet 4.5) |
| 13 | Qwen 3 14B (Non-reasoning) | 33.33 | #537 (Qwen 3 14B) |
| 14 | Gemma 3 12B (IT) | 33.26 | #664 |
| 15 | Bielik-11B-v2.3-Instruct | 33.25 | #693 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=euroeval-hungarian-summarization-hunsum · How It Works · Data refreshed daily, snapshot 2026-10-20.