Falcon2-5.5B-multilingual: benchmark results
A community prune by ssmits of TII's Falcon-11B, removing 50% of layers via the PruneMe method analyzed across 11 languages, requiring further continued pretraining to restore full performance. Provider: ssmits. Released 2024-05-21. Access: Open.
Aggregate Rating 1063 ± 38, rank #1628 of 1635 rated models, from 29 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| Open LLM Leaderboard v1 - TruthfulQA MC2 | 47.56 | MC2 (%) (0-shot) | 38.1 |
| EuroEval Icelandic Knowledge | 0.83 | Knowledge Average Score (%) | 17.2 |
| EuroEval Danish Knowledge | 4.89 | Knowledge Average Score (%) | 15.7 |
| EuroEval Norwegian Knowledge - Idioms NO | -0.36 | MCC (x100) | 14.1 |
| EuroEval Danish Knowledge - Danish Citizen Tests | 4.89 | MCC (x100) | 12.6 |
| EuroEval Portuguese Common Sense Reasoning | 0.16 | Common Sense Reasoning Average Score (%) | 10.7 |
| Open LLM Leaderboard v1 - WinoGrande | 55.25 | Accuracy (%) (5-shot) | 9 |
| Open LLM Leaderboard v1 - HellaSwag | 40.3 | Normalized accuracy (%) (10-shot) | 8.1 |
| Open LLM Leaderboard v1 - ARC Challenge | 26.11 | Normalized accuracy (%) (25-shot) | 5.4 |
| Open LLM Leaderboard v1 - GSM8K | 0 | Accuracy (%) (5-shot) | 4.5 |
| EuroEval Portuguese NLU - MultiWikiQA PT | 0.2 | Character F1 (%) | 4.1 |
| EuroEval Norwegian Knowledge - NRK Quiz QA | 0.1 | MCC (x100) | 3.9 |
Interactive version: theaggregate.ai/model?slug=falcon2-5-5b-multilingual · How It Works · Data refreshed daily, snapshot 2026-11-05.