EVALITA - evalita NER: leaderboard
The "evalita NER" subtask of EVALITA. A task from EVALITA, the periodic evaluation campaign for Italian-language natural language processing and understanding.
Metric: CPS. Source: huggingface.co. Saturation forecast: Around 2030. 48 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Llama 3.3 70B Instruct | 44.58 | #534 |
| 2 | Llama 3.1 8B Instruct | 40.3 | #1035 |
| 3 | Llama-3.1-SuperNova-Lite | 39.49 | #730 |
| 4 | Llama-3.1-Tulu-3-70B | 39.12 | #482 |
| 5 | Qwen 2.5 72B Instruct | 38.91 | #454 |
| 6 | Mistral Large 2 (Nov) Instruct (2411) | 38.79 | #465 |
| 7 | Gemma 3 27B (IT) | 38.76 | #523 |
| 8 | calme-3.2-instruct-78B | 38.52 | #347 |
| 9 | Gemma 2 27B (IT) | 38.33 | #645 |
| 10 | Mistral Small 3 | 38.16 | #627 |
| 11 | Gemma 2 9B (IT) | 38.03 | #788 |
| 12 | Gemma 3 12B (IT) | 37.86 | #664 |
| 13 | Qwen 2.5 VL 32B Instruct | 37.81 | #453 |
| 14 | DeepSeek R1 Distill Llama 70B | 37.76 | #594 |
| 15 | Gemma 3 4B (IT) | 36 | #992 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=evalita-evalita-ner · How It Works · Data refreshed daily, snapshot 2026-10-20.