family phi · 1 variants · unknown
| benchmark | category | best score |
|---|---|---|
| ARC-Challenge | knowledge | 90.7 |
| GSM8K | math | 89.6 |
| OpenBookQA | knowledge | 88.0 |
| PIQA | knowledge | 86.9 |
| BoolQ | knowledge | 84.8 |
| WinoGrande | knowledge | 81.5 |
| CommonsenseQA | knowledge | 80.0 |
| BBH | reasoning | 79.1 |
| HellaSwag | knowledge | 77.0 |
| MMLU | knowledge | 75.7 |
| MBPP | coding | 71.7 |
| TruthfulQA | safety | 70.2 |
| MedQA | domain | 65.4 |
| HumanEval | coding | 61.0 |
| TriviaQA | knowledge | 58.1 |
| MATH | math | 51.6 |
| AGIEval | reasoning | 45.1 |
| variant | kind | quant | bits | vRAM | dl | org | year |
|---|---|---|---|---|---|---|---|
| Phi-3-small-8k-instruct | original | bf16 | 16.0 | 15,205 | microsoft | 2025 |