openchat-3.5-0106-laser: benchmark results
QuixiAI's laser-pruned version of openchat/openchat-3.5-0106, applying laserRMT to identify and adjust high-noise layers. Provider: Cognitive Computations. Released 2024-01-27. Access: Open.
Aggregate Rating 1429 ± 16, rank #1103 of 1656 rated models, from 29 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| Open Korean LLM - Ko-IFEval | 43.5 | Strict accuracy, prompt/instruction mean (%) | 87.9 |
| Open LLM Leaderboard v1 - GSM8K | 68.92 | Accuracy (%) (5-shot) | 87.3 |
| Open Korean LLM - Ko-EQ-Bench | 35.12 | EQ-Bench score | 85 |
| Open Portuguese LLM - FaQuAD NLI | 78.67 | Macro F1 (%) | 84.9 |
| Open Korean LLM - KorNAT-Helpful | 49.8 | Normalized accuracy (%) | 81.9 |
| Open LLM Leaderboard v1 - MMLU | 65.11 | Accuracy (%) (5-shot) | 81.9 |
| Open Korean LLM Leaderboard | 41.13 | Average Score (%) | 80.9 |
| Open Ko-LLM Leaderboard | 41.13 | Average (%) | 80.8 |
| Open Portuguese LLM - ASSIN2 RTE | 92.48 | Macro F1 (%) | 79.5 |
| Open LLM Leaderboard v1 - WinoGrande | 81.45 | Accuracy (%) (5-shot) | 79.3 |
| Open Korean LLM - Ko-GSM8K (flexible extract) | 39.95 | Exact match, flexible extract (%) | 74.4 |
| Open LLM Leaderboard v1 - ARC Challenge | 66.04 | Normalized accuracy (%) (25-shot) | 72.1 |
Interactive version: theaggregate.ai/model?slug=openchat-3-5-0106-laser · How It Works · Data refreshed daily, snapshot 2026-10-20.