Lexora-Lite-3B: benchmark results
DeepMount00's Italian instruction fine-tune of Qwen2.5 3B, trained with SFT and DPO on Sonnet-3.5-generated Italian data. Provider: Other. Released 2024-09-19. Access: Open.
Aggregate Rating 1458 ± 37, rank #3444 of 10916 rated models, from 13 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| Pinocchio Italian - Lingua Straniera | 78.45 | Accuracy (%) | 93.2 |
| Pinocchio Italian - Logica | 47.9 | Accuracy (%) | 88.6 |
| Pinocchio Italian - Matematica E Scienze | 58.21 | Accuracy (%) | 88.6 |
| Pinocchio Italian Leaderboard | 60.64 | Average Accuracy (%) | 88.6 |
| Pinocchio Italian - Generale | 60.61 | Accuracy (%) | 86.4 |
| Pinocchio Italian - Diritto | 56.92 | Accuracy (%) | 84.1 |
| Open LLM Leaderboard - MATH Level 5 | 23.04 | Score | 76.4 |
| Pinocchio Italian - Cultura | 61.73 | Accuracy (%) | 75 |
| Open LLM Leaderboard - IFEval | 57.76 | Score | 69.8 |
| Open LLM Leaderboard - MMLU-Pro | 36.02 | Score | 54.9 |
| Open LLM Leaderboard - BBH | 48.73 | Score | 43.4 |
| Open LLM Leaderboard - MuSR | 39.66 | Score | 39.2 |
Interactive version: theaggregate.ai/model?slug=lexora-lite-3b · How It Works · Data refreshed daily, snapshot 2026-11-08.