Llama 3.2 3B: benchmark results
Meta's Llama 3.2 3B, the base pretrained checkpoint in Meta's Llama 3.2 family released September 2024; the linked model card provides no further details here. Provider: Meta. Released 2024-09-25. Access: Open.
Aggregate Rating 1344 ± 9, rank #1396 of 1656 rated models, from 420 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| aiXamine - Model & Data Privacy | 88.1 | Service score (0-100; normalized pass rate) | 97.6 |
| LA Leaderboard - Summarization Galician | 7.42 | BLEU | 97 |
| LA Leaderboard - RagQuAS | 81.1 | Semantic Answer Similarity (x100) | 94.1 |
| LA Leaderboard - caBREU Catalan | 16.88 | BLEU | 86.8 |
| aiXamine - Over-Refusal | 93.8 | Service score (0-100; normalized pass rate) | 85.4 |
| EuroEval Finnish NLU - Tydiqa FI | 70.5 | Character F1 (%) | 84.7 |
| Polar (US) - Sociocultural Axis | 84.66 | Sociocultural-axis ICAT (0-100), the StereoSet idealized con | 83.8 |
| LA Leaderboard - EusReading Basque | 32.95 | Accuracy (%) | 83.1 |
| LA Leaderboard - BHTCv2 Basque | 29.45 | F1 (%) | 82.4 |
| LA Leaderboard - CoQCat | 70 | F1 (%) | 81.5 |
| LA Leaderboard - OpenBookQA Galician | 27.6 | Accuracy (%) | 80.9 |
| LA Leaderboard - XCOPA Basque | 53.4 | Accuracy (%) | 80.9 |
Interactive version: theaggregate.ai/model?slug=llama-3-2-3b · How It Works · Data refreshed daily, snapshot 2026-10-20.