openchat_3.5: benchmark results
The OpenChat team's C-RLFT fine-tune of Mistral 7B on mixed-quality public data, an early 7B claiming ChatGPT-comparable MT-Bench scores. Provider: OpenChat. Released 2023-11-01. Access: Open.
Aggregate Rating 1430 ± 17, rank #1098 of 1656 rated models, from 41 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| Open Korean LLM - Ko-EQ-Bench | 33.21 | EQ-Bench score | 83 |
| Open Korean LLM - Ko-IFEval | 40.28 | Strict accuracy, prompt/instruction mean (%) | 82.7 |
| Open Korean LLM Leaderboard | 41.19 | Average Score (%) | 81.7 |
| Open Ko-LLM Leaderboard | 41.19 | Average (%) | 81.5 |
| Open Korean LLM - Ko-GSM8K (flexible extract) | 42.46 | Exact match, flexible extract (%) | 77.3 |
| Open LLM Leaderboard - IFEval | 59.31 | Score | 71.1 |
| Open Korean LLM - KorNAT-Helpful | 48.25 | Normalized accuracy (%) | 66.3 |
| Open LLM Leaderboard - MuSR | 42.29 | Score | 63.3 |
| Open LLM Leaderboard - GPQA | 29.87 | Score | 54.6 |
| BenchBench | 52.71 | Aggregate Score (%) | 54.4 |
| YALL Nous Leaderboard | 52.7 | Average | 50 |
| Open Korean LLM - KorNAT-Harmless | 62.45 | Normalized accuracy (%) | 44.1 |
Interactive version: theaggregate.ai/model?slug=openchat-3-5 · How It Works · Data refreshed daily, snapshot 2026-10-20.