KoSOLAR-10.7B-QLoRA-NEFTune-kolon-v2.0: benchmark results

Provider: Other. Access: Open.

Unified ELO 1379 ± 18, rank #2057 of 2656 rated models, from 10 benchmark results.

Strongest benchmark results

BenchmarkScoreMetricPercentile
Open Korean LLM - Ko-GPQA-Diamond25.76Normalized accuracy (%)81.3
Open Korean LLM - KorNAT-Helpful49.08Normalized accuracy (%)76.2
Open Korean LLM - Ko-Winogrande60.38Accuracy (%)53.7
Open Korean LLM - Ko-IFEval27.79Strict accuracy, prompt/instruction mean (%)45.2
Open Korean LLM Leaderboard32.44Average Score (%)33.6
Open Ko-LLM Leaderboard32.44Average (%)33.3
Open Korean LLM - Ko-EQ-Bench0EQ-Bench score24.5
Open Korean LLM - Ko-GSM8K (flexible extract)1.97Exact match, flexible extract (%)16.9
Open Korean LLM - KorNAT-Harmless60.32Normalized accuracy (%)13.7
Open Korean LLM - KorNAT-CKA15.51Normalized accuracy (%)9.3

Interactive version: theaggregate.ai/model?slug=kosolar-10-7b-qlora-neftune-kolon-v2-0 · How It Works · Data refreshed daily, snapshot 2026-09-19.