EVALITA LLM Leaderboard: leaderboard
EVALITA — the Italian NLP evaluation campaign's LLM leaderboard; aggregate accuracy over Italian-language understanding and generation tasks.
Metric: Average CPS. Source: huggingface.co. Saturation forecast: Around July 2028. 48 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Mistral Large 2 (Nov) Instruct (2411) | 62.28 | #465 |
| 2 | Llama-3.1-Tulu-3-70B | 60.16 | #482 |
| 3 | Gemma 3 27B (IT) | 59.2 | #523 |
| 4 | Llama 3.3 70B Instruct | 57.93 | #534 |
| 5 | Mistral Small 3 | 57.67 | #627 |
| 6 | Qwen 2.5 72B Instruct | 57.36 | #454 |
| 7 | Qwen 2.5 VL 32B Instruct | 57.26 | #453 |
| 8 | calme-3.2-instruct-78B | 56.82 | #347 |
| 9 | Qwen 3 Next 80B A3B Instruct | 55.35 | #368 |
| 10 | Qwen2.5-14B-Instruct-1M | 54.72 | #578 |
| 11 | DeepSeek R1 Distill Llama 70B | 54.66 | #594 |
| 12 | Gemma 3 12B (IT) | 53.92 | #664 |
| 13 | Gemma 2 27B (IT) | 53.86 | #645 |
| 14 | Gemma 2 9B (IT) | 53.64 | #788 |
| 15 | Qwen 2.5 7B Instruct | 52.62 | #862 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=evalita-llm-leaderboard · How It Works · Data refreshed daily, snapshot 2026-10-20.