Liberated-Qwen1.5-14B: benchmark results
Abacus.AI's Liberated-Qwen1.5-14B, an uncensored fine-tune of Qwen1.5-14B at 8K sequence length, trained on the SystemChat dataset for system-prompt compliance across long multi-turn conversations. Provider: Abacus.AI. Released 2024-04-01. Access: Open.
Aggregate Rating 1430 ± 18, rank #1097 of 1656 rated models, from 32 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| Open Medical LLM - MMLU College Medicine | 68.21 | Accuracy (%) | 87 |
| Open Korean LLM - Ko-GSM8K (flexible extract) | 49.58 | Exact match, flexible extract (%) | 83.3 |
| Open Ko-LLM Leaderboard | 41.37 | Average (%) | 82.4 |
| Open Korean LLM Leaderboard | 41.37 | Average Score (%) | 82.4 |
| Open Korean LLM - Ko-EQ-Bench | 32.75 | EQ-Bench score | 82.1 |
| Open Portuguese LLM - ASSIN2 RTE | 92.52 | Macro F1 (%) | 80 |
| Open Medical LLM - MMLU Clinical Knowledge | 73.96 | Accuracy (%) | 69.3 |
| Open Medical LLM - MMLU Professional Medicine | 72.79 | Accuracy (%) | 66.4 |
| Open LLM Leaderboard - MATH Level 5 | 16.01 | Score | 63.8 |
| Open Korean LLM - Ko-IFEval | 32.59 | Strict accuracy, prompt/instruction mean (%) | 63.2 |
| Open Medical LLM - MMLU Anatomy | 64.44 | Accuracy (%) | 62.8 |
| Open Korean LLM - KorNAT-Helpful | 47.65 | Normalized accuracy (%) | 58.3 |
Interactive version: theaggregate.ai/model?slug=liberated-qwen1-5-14b · How It Works · Data refreshed daily, snapshot 2026-10-20.