Phi-3-small-128k-instruct — benchmark results
Microsoft's 7B Phi-3 Small instruct model with a 128K context, trained on synthetic and filtered web data. Provider: Microsoft. Released 2024-05-07. Access: Open.
Unified ELO 1450 ± 16, rank #1000 of 1776 rated models, from 32 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| Open LLM Leaderboard - BBH | 45.63 | Score | 85.7 |
| Open LLM Leaderboard - MMLU-Pro | 38.78 | Score | 83.5 |
| Open LLM Leaderboard - MuSR | 14.5 | Score | 79.2 |
| Open LLM Leaderboard - IFEval | 63.68 | Score | 75.9 |
| Open LLM Leaderboard - GPQA | 8.95 | Score | 73.2 |
| Open LLM Leaderboard - MATH Level 5 | 20.26 | Score | 71.4 |
| BenchBench | 65.61 | Aggregate Score (%) | 69.9 |
| LiveBench AMPS Hard | 28 | Score | 63.2 |
| LiveBench Web Of Lies V2 | 28 | Score | 58.3 |
| ChineseSafe Benchmark | 67.43 | Accuracy (%) | 55.6 |
| LiveBench Coding Completion | 19.74 | Score | 55.6 |
| LiveBench LCB Generation | 30 | Score | 53.5 |
Interactive version: theaggregate.ai/model?slug=phi-3-small-128k-instruct · How the rankings work · Data refreshed daily, snapshot 2026-07-22.