Phi-3-medium-128k-instruct: benchmark results
Microsoft's 14B Phi-3 Medium instruct model with a 128K context, trained on synthetic and filtered web data. Provider: Microsoft. Released 2024-05-21. Access: Open.
Aggregate Rating 1448 ± 12, rank #1013 of 1656 rated models, from 126 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| ARC Challenge (AI2) | 91.6 | Accuracy (%) | 94.7 |
| Big-Bench Hard | 81.4 | Average (%) | 93.8 |
| OpenBookQA | 87.4 | Accuracy (%) | 92.9 |
| Open Korean LLM - Ko-EQ-Bench | 42.57 | EQ-Bench score | 92.4 |
| Open Portuguese LLM - ASSIN2 RTE | 93.83 | Macro F1 (%) | 92.4 |
| Open LLM Leaderboard - BBH | 63.82 | Score | 88.1 |
| BBH | 81.4 | Score (imported) | 86.7 |
| Open LLM Leaderboard - MMLU-Pro | 47.12 | Score | 86 |
| Open LLM Leaderboard - GPQA | 33.64 | Score | 83.1 |
| Open Portuguese LLM - ENEM | 71.73 | Accuracy (%) | 81.3 |
| Open Portuguese LLM - BLUEX | 61.47 | Accuracy (%) | 80.6 |
| Open Portuguese LLM - OAB Exams | 51.44 | Accuracy (%) | 80 |
Interactive version: theaggregate.ai/model?slug=phi-3-medium-128k-instruct · How It Works · Data refreshed daily, snapshot 2026-10-20.