open-cabrita3B: benchmark results

Provider: Other. Access: Open.

Unified ELO 1216 ± 20, rank #2876 of 2928 rated models, from 11 benchmark results.

Strongest benchmark results

BenchmarkScoreMetricPercentile
Open Portuguese LLM - FaQuAD NLI43.97Macro F1 (%)17.1
Open LLM Leaderboard v1 - GSM8K0.99Accuracy (%) (5-shot)15.1
Open Portuguese LLM - ASSIN2 RTE43.01Macro F1 (%)13.5
Open LLM Leaderboard v1 - WinoGrande59.43Accuracy (%) (5-shot)12.6
Open LLM Leaderboard v1 - HellaSwag55.35Normalized accuracy (%) (10-shot)12.3
Open LLM Leaderboard v1 - ARC Challenge33.79Normalized accuracy (%) (25-shot)11.9
Open Portuguese LLM - BLUEX21.14Accuracy (%)11.4
Open LLM Leaderboard v1 - TruthfulQA MC238.5MC2 (%) (0-shot)8.3
Open LLM Leaderboard v1 - MMLU25.16Accuracy (%) (5-shot)6.8
Open Portuguese LLM - OAB Exams22.69Accuracy (%)6.7
Open Portuguese LLM - ENEM17.98Accuracy (%)5

Interactive version: theaggregate.ai/model?slug=open-cabrita3b · How It Works · Data refreshed daily, snapshot 2026-09-23.