Phi-1: benchmark results
Microsoft's Phi-1, a 1.3B transformer specialized for basic Python coding, trained on curated Python code from The Stack, StackOverflow Q&A and synthetic textbooks generated by GPT-3.5. Provider: Microsoft. Released 2023-06-20. Access: Open.
Aggregate Rating 1125 ± 25, rank #1639 of 1656 rated models, from 24 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| AI Energy Score (Text Generation) | 5 | Energy Score (1-5) | 79.5 |
| BigCode Models Leaderboard | 51.2 | HumanEval Python Pass@1 (%) | 62.7 |
| Open Portuguese LLM - BLUEX | 24.9 | Accuracy (%) | 19.6 |
| Open LLM Leaderboard - GPQA | 26.51 | Score | 19.3 |
| Open Portuguese LLM - FaQuAD NLI | 43.97 | Macro F1 (%) | 17.1 |
| Open LLM Leaderboard - MuSR | 35.25 | Score | 13.9 |
| Open LLM Leaderboard - IFEval | 20.68 | Score | 12.9 |
| Open Portuguese LLM - OAB Exams | 23.69 | Accuracy (%) | 10.6 |
| Open Portuguese LLM - ASSIN2 RTE | 34.47 | Macro F1 (%) | 9.2 |
| Open LLM Leaderboard - BBH | 31.39 | Score | 9 |
| LLMZSZL Leaderboard | 25.73 | Score | 8.2 |
| Open LLM Leaderboard - MMLU-Pro | 11.62 | Score | 7.4 |
Interactive version: theaggregate.ai/model?slug=phi-1 · How It Works · Data refreshed daily, snapshot 2026-10-20.