Orca-2-13B [bfloat16] [rev: 2539ff53e6ba]: benchmark results
Microsoft's Orca 2 research model, fine-tuned from Llama 2 on synthetic reasoning data for single-turn reasoning, reading comprehension, mathematics and summarization. It was not optimized for chat or trained with RLHF/DPO; this row preserves its historical revision and evaluation precision. Provider: Microsoft. Access: Open.
Aggregate Rating 1388 ± 32, rank #6598 of 10916 rated models, from 6 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| Open LLM Leaderboard v1 - TruthfulQA MC2 | 56.42 | MC2 (%) (0-shot) | 67.9 |
| Open LLM Leaderboard v1 - GSM8K | 37.83 | Accuracy (%) (5-shot) | 52.9 |
| Open LLM Leaderboard v1 - ARC Challenge | 60.92 | Normalized accuracy (%) (25-shot) | 50.3 |
| Open LLM Leaderboard v1 - WinoGrande | 76.56 | Accuracy (%) (5-shot) | 49 |
| Open LLM Leaderboard v1 - MMLU | 60.3 | Accuracy (%) (5-shot) | 47.6 |
| Open LLM Leaderboard v1 - HellaSwag | 79.85 | Normalized accuracy (%) (10-shot) | 39.4 |
Interactive version: theaggregate.ai/model?slug=orca-2-13b-bfloat16-rev-2539ff53e6ba · How It Works · Data refreshed daily, snapshot 2026-11-08.