Phi-3-medium-128k-instruct [bfloat16] [rev: 1e10cf49da9e]: benchmark results
Microsoft's Phi-3 Medium instruction model, trained on synthetic and filtered web data and post-trained with supervised fine-tuning and direct preference optimization. This row preserves its context variant and separately evaluated prompt, revision or precision configuration. Provider: Microsoft. Access: Open.
Aggregate Rating 1484 ± 40, rank #2608 of 10916 rated models, from 6 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| Open LLM Leaderboard v1 - GSM8K | 81.05 | Accuracy (%) (5-shot) | 99.8 |
| Open LLM Leaderboard v1 - MMLU | 76.68 | Accuracy (%) (5-shot) | 97.9 |
| Open LLM Leaderboard v1 - ARC Challenge | 66.47 | Normalized accuracy (%) (25-shot) | 73.7 |
| Open LLM Leaderboard v1 - HellaSwag | 84.85 | Normalized accuracy (%) (10-shot) | 69.3 |
| Open LLM Leaderboard v1 - TruthfulQA MC2 | 54.52 | MC2 (%) (0-shot) | 62.8 |
| Open LLM Leaderboard v1 - WinoGrande | 74.43 | Accuracy (%) (5-shot) | 36.5 |
Interactive version: theaggregate.ai/model?slug=phi-3-medium-128k-instruct-bfloat16-rev-1e10cf49da9e · How It Works · Data refreshed daily, snapshot 2026-11-08.