EVALITA - faq: leaderboard
The "faq" 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: Not forecast. 48 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | LLaMAntino-3-ANITA-8B-Inst-DPO-ITA | 56.55 | #1009 |
| 2 | Llama-3.1-SuperNova-Lite | 54.89 | #730 |
| 3 | granite-3.1-8B-instruct | 54.7 | #1069 |
| 4 | Mistral Large 2 (Nov) Instruct (2411) | 54.69 | #465 |
| 5 | Mistral Small 3 | 54.55 | #627 |
| 6 | Llama-3.1-Tulu-3-70B | 54.52 | #482 |
| 7 | Qwen2.5-14B-Instruct-1M | 54.41 | #578 |
| 8 | aya-expanse-8B | 54.39 | #931 |
| 9 | Llama 3.3 70B Instruct | 54.3 | #534 |
| 10 | Phi-4 | 54.3 | #706 |
| 11 | Qwen 2.5 72B Instruct | 54 | #454 |
| 12 | Llama 4 Scout Instruct | 53.97 | #539 |
| 13 | Qwen 2.5 VL 32B Instruct | 53.84 | #453 |
| 14 | Pixtral-12B | 53.83 | #808 |
| 15 | Llama 3.1 8B Instruct | 53.73 | #1035 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=evalita-faq · How It Works · Data refreshed daily, snapshot 2026-10-20.