EVALITA - sentiment-analysis: leaderboard
The "sentiment-analysis" 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 January 2027. 48 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Mistral Large 2 (Nov) Instruct (2411) | 80.94 | #465 |
| 2 | Llama-3.1-Tulu-3-70B | 80.41 | #482 |
| 3 | Gemma 3 27B (IT) | 80.3 | #523 |
| 4 | calme-3.2-instruct-78B | 80.15 | #347 |
| 5 | Llama 3.3 70B Instruct | 79.55 | #534 |
| 6 | Gemma 3 12B (IT) | 78.34 | #664 |
| 7 | Qwen 2.5 72B Instruct | 77.68 | #454 |
| 8 | DeepSeek R1 Distill Llama 70B | 77.48 | #594 |
| 9 | Gemma 3n E4B (IT) | 76.34 | #749 |
| 10 | Mistral Small 3 | 75.95 | #627 |
| 11 | Gemma 3 4B (IT) | 75.3 | #992 |
| 12 | Llama-3.1-SuperNova-Lite | 75.24 | #730 |
| 13 | Phi-4 | 75.21 | #706 |
| 14 | granite-3.1-8B-instruct | 74.76 | #1069 |
| 15 | Gemma 2 27B (IT) | 74.26 | #645 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=evalita-sentiment-analysis · How It Works · Data refreshed daily, snapshot 2026-10-20.