EVALITA - word-in-context: leaderboard
The "word-in-context" 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 June 2027. 48 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Qwen 2.5 VL 32B Instruct | 77.56 | #453 |
| 2 | Mistral Large 2 (Nov) Instruct (2411) | 75.55 | #465 |
| 3 | calme-3.2-instruct-78B | 73.44 | #347 |
| 4 | Qwen 2.5 72B Instruct | 72.06 | #454 |
| 5 | Mistral Small 3 | 71.65 | #627 |
| 6 | Qwen 3 Next 80B A3B Instruct | 71.41 | #368 |
| 7 | DeepSeek R1 Distill Llama 70B | 69.26 | #594 |
| 8 | Phi-3-medium-4k-instruct | 68.92 | #952 |
| 9 | Gemma 2 27B (IT) | 68.56 | #645 |
| 10 | Llama 3.3 70B Instruct | 67.15 | #534 |
| 11 | Gemma 3 12B (IT) | 66.5 | #664 |
| 12 | Gemma 3 27B (IT) | 66.49 | #523 |
| 13 | Llama 3 8B 4bit UltraChat Ita | 66.22 | #1280 |
| 14 | Llama-3.1-Tulu-3-70B | 65.44 | #482 |
| 15 | Phi-3.5-mini-instruct | 65.33 | #1167 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=evalita-word-in-context · How It Works · Data refreshed daily, snapshot 2026-10-20.