Mistral-7B-OpenOrca: benchmark results
Open-Orca's fine-tune of Mistral 7B on their GPT-4-augmented OpenOrca dataset, reproducing the Microsoft Orca paper's approach, trained with Axolotl and an 8K context. Provider: Open-Orca. Released 2023-09-29. Access: Open.
Aggregate Rating 1348 ± 15, rank #1389 of 1656 rated models, from 43 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| LLM Trustworthy - Out-of-Distribution | 73.4 | Trust Score (%) | 75 |
| Open Korean LLM - KorNAT-Helpful | 48.88 | Normalized accuracy (%) | 74.1 |
| Open LLM Leaderboard v1 - ARC Challenge | 64.08 | Normalized accuracy (%) (25-shot) | 65.4 |
| Open LLM Leaderboard v1 - HellaSwag | 83.99 | Normalized accuracy (%) (10-shot) | 63 |
| Open Korean LLM - Ko-GSM8K (flexible extract) | 28.51 | Exact match, flexible extract (%) | 62.9 |
| Open Korean LLM - Ko-EQ-Bench | 12.9 | EQ-Bench score | 58.7 |
| Open LLM Leaderboard - IFEval | 49.78 | Score | 58.5 |
| LLM Trustworthy - Adversarial | 47.23 | Trust Score (%) | 58.3 |
| Open LLM Leaderboard v1 - TruthfulQA MC2 | 53.05 | MC2 (%) (0-shot) | 58.3 |
| Open LLM Leaderboard v1 - WinoGrande | 77.74 | Accuracy (%) (5-shot) | 58.2 |
| Open Korean LLM - Ko-IFEval | 30.58 | Strict accuracy, prompt/instruction mean (%) | 56.8 |
| Open LLM Leaderboard v1 - MMLU | 62.24 | Accuracy (%) (5-shot) | 54.6 |
Interactive version: theaggregate.ai/model?slug=mistral-7b-openorca · How It Works · Data refreshed daily, snapshot 2026-10-20.