phi-2 [bfloat16] [rev: b10c3eba545a]: benchmark results
Microsoft's Phi-2 base transformer, trained on synthetic NLP text and filtered websites without RLHF fine-tuning. This row preserves its separately evaluated revision, precision, shot count or runtime configuration. Provider: Microsoft. Access: Open.
Aggregate Rating 1372 ± 34, rank #7103 of 10916 rated models, from 6 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| Open LLM Leaderboard v1 - GSM8K | 54.97 | Accuracy (%) (5-shot) | 66.6 |
| Open LLM Leaderboard v1 - ARC Challenge | 61.01 | Normalized accuracy (%) (25-shot) | 50.8 |
| Open LLM Leaderboard v1 - MMLU | 57.92 | Accuracy (%) (5-shot) | 42.2 |
| Open LLM Leaderboard v1 - WinoGrande | 73.48 | Accuracy (%) (5-shot) | 32.3 |
| Open LLM Leaderboard v1 - TruthfulQA MC2 | 44.24 | MC2 (%) (0-shot) | 26.4 |
| Open LLM Leaderboard v1 - HellaSwag | 74.92 | Normalized accuracy (%) (10-shot) | 25.5 |
Interactive version: theaggregate.ai/model?slug=phi-2-bfloat16-rev-b10c3eba545a · How It Works · Data refreshed daily, snapshot 2026-11-08.