openchat-3.5-0106: benchmark results
OpenChat's openchat-3.5-0106, a Mistral 7B fine-tune with separate Coding+Generalist and Mathematical Reasoning modes, plus experimental evaluator and feedback capabilities. Provider: OpenChat. Released 2024-01-06. Access: Open.
Aggregate Rating 1424 ± 11, rank #1130 of 1656 rated models, from 115 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| Open PL LLM - KLEJ NER (generative, 5-shot) | 56.37 | Accuracy (%) | 91.4 |
| Open Korean LLM - Ko-IFEval | 44.1 | Strict accuracy, prompt/instruction mean (%) | 88.4 |
| Open Portuguese LLM - FaQuAD NLI | 79.21 | Macro F1 (%) | 86.7 |
| Open PL LLM - KLEJ NER (multiple choice, 5-shot) | 55.44 | Accuracy (%) | 85.9 |
| Open LLM Leaderboard v1 - GSM8K | 68.16 | Accuracy (%) (5-shot) | 85.2 |
| Open PL LLM - PoQuAD Open Book (generative, 5-shot) | 70.09 | Levenshtein Similarity (%) | 85.2 |
| Open Korean LLM - Ko-EQ-Bench | 35.03 | EQ-Bench score | 84.7 |
| Open Portuguese LLM - ASSIN2 RTE | 92.73 | Macro F1 (%) | 82.8 |
| Open Korean LLM Leaderboard | 41.14 | Average Score (%) | 81 |
| Open Ko-LLM Leaderboard | 41.14 | Average (%) | 80.9 |
| Open LLM Leaderboard v1 - MMLU | 65.04 | Accuracy (%) (5-shot) | 80.8 |
| Open LLM Leaderboard v1 - WinoGrande | 81.77 | Accuracy (%) (5-shot) | 80.8 |
Interactive version: theaggregate.ai/model?slug=openchat-3-5-0106 · How It Works · Data refreshed daily, snapshot 2026-10-20.