Pinocchio Italian - Matematica E Scienze: leaderboard
The "Matematica E Scienze" subtask of Pinocchio Italian. A task from the Pinocchio Italian benchmark, probing factual knowledge and truthfulness of language models on Italian-language questions.
Metric: Accuracy (%). Source: huggingface.co. Saturation forecast: Estimated already saturated. 45 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Lexora-Medium-7B | 68.66 | #773 |
| 2 | Gemma 2 27B (IT) | 67.68 | #645 |
| 3 | Gemma 2 9B (IT) | 65.64 | #788 |
| 4 | Phi-3-medium-4k-instruct | 64.17 | #952 |
| 5 | Llama 3.1 8B Ita | 52.65 | #797 |
| 6 | Llama 3.1 8B Instruct | 51.34 | #1035 |
| 7 | Volare | 50.96 | #1213 |
| 8 | Llama 3 8B Ita | 49.42 | #926 |
| 9 | Llama 3 8B Instruct | 48.69 | #1129 |
| 10 | LLaMAntino-3-ANITA-8B-Inst-DPO-ITA | 48.36 | #1009 |
| 11 | Gemma 2 2B (IT) | 46.32 | #1341 |
| 12 | Qwen 2 1.5B Instruct | 43.06 | #1426 |
| 13 | Llama 3 8B 4bit UltraChat Ita | 42.4 | #1280 |
| 14 | Mistral-7B-v0.1 | 40.99 | #1326 |
| 15 | Mistral-7B-v0.3 | 39.58 | #1355 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=pinocchio-italian-matematica-e-scienze · How It Works · Data refreshed daily, snapshot 2026-10-20.