EVALITA - text-entailment: leaderboard
The "text-entailment" 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 | Qwen 2.5 72B Instruct | 85.07 | #454 |
| 2 | Qwen2.5-14B-Instruct-1M | 85.02 | #578 |
| 3 | calme-3.2-instruct-78B | 84.31 | #347 |
| 4 | Qwen 2.5 VL 32B Instruct | 83.7 | #453 |
| 5 | Qwen 2.5 7B Instruct | 83.65 | #862 |
| 6 | Llama-3.1-Tulu-3-70B | 82.42 | #482 |
| 7 | Mistral Small 3 | 82 | #627 |
| 8 | Gemma 3 27B (IT) | 81.13 | #523 |
| 9 | Mistral Large 2 (Nov) Instruct (2411) | 80.96 | #465 |
| 10 | Llama 3.3 70B Instruct | 80.79 | #534 |
| 11 | Gemma 2 9B (IT) | 79.97 | #788 |
| 12 | GLM-4.1V-9B (Thinking) | 79.79 | #472 (GLM-4.1V-9B) |
| 13 | Gemma 3 12B (IT) | 79.52 | #664 |
| 14 | DeepSeek R1 Distill Llama 70B | 79.37 | #594 |
| 15 | Gemma 2 27B (IT) | 78.72 | #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-text-entailment · How It Works · Data refreshed daily, snapshot 2026-10-20.