oxyge1-33B: benchmark results

Sakalti's TIES merge of QwQ-32B-Preview and Qwen2.5 32B Instruct built with mergekit. Provider: Sakalti. Released 2024-12-08. Access: Open.

Aggregate Rating 1528 ± 26, rank #690 of 1656 rated models, from 184 benchmark results.

Strongest benchmark results

BenchmarkScoreMetricPercentile
Open LLM Leaderboard - MMLU-Pro59.09Score99.7
Open LLM Leaderboard - BBH70.33Score99.1
Open LLM Leaderboard - MuSR50.08Score98.4
Open Portuguese LLM - ASSIN2 RTE94.69Macro F1 (%)98
Open Japanese LLM - Wiki PAS SET F111.91Score (%)97.6
Open Portuguese LLM - BLUEX78.3Accuracy (%)97.4
Open Portuguese LLM - ENEM84.39Accuracy (%)97.4
Open LLM Leaderboard - MATH Level 549.62Score96.7
Open Portuguese LLM - OAB Exams63.33Accuracy (%)95.8
Open LLM Leaderboard - GPQA38.26Score95.6
Open Japanese LLM - Jnli Exact Match89.69Score (%)95.1
Open Japanese LLM - Jsick Exact Match85.73Score (%)95

Interactive version: theaggregate.ai/model?slug=oxyge1-33b · How It Works · Data refreshed daily, snapshot 2026-10-20.