Microsoft: AI model benchmarks
AI models from Microsoft. 24 models tracked, 22 with a unified ELO rating. Current best: Phi-Line_14B (ELO 1546, rank #454 overall). Most recent tracked release: 2025-04-30.
Rated models
| Rank | Model | Provider | Unified ELO | Benchmarks |
|---|---|---|---|---|
| 454 | Phi-Line_14B | Microsoft | 1546 ± 1 | 9 |
| 621 | phi-4-14B | Microsoft | 1516 ± 1 | 17 |
| 627 | Phi-lthy4 | Microsoft | 1514 ± 1 | 9 |
| 636 | Phi-3.5-MoE-instruct | Microsoft | 1512 ± 1 | 63 |
| 641 | WizardLM-2 8x22B | Microsoft | 1511 ± 1 | 52 |
| 692 | Phi-4 | Microsoft | 1497 ± 1 | 705 |
| 794 | Phi-3-medium-4k-instruct | Microsoft | 1477 ± 1 | 122 |
| 816 | Phi-3-small-128k-instruct | Microsoft | 1474 ± 1 | 21 |
| 820 | Phi-3-medium-128k-instruct | Microsoft | 1473 ± 1 | 60 |
| 855 | Phi-3-small-8k-instruct | Microsoft | 1467 ± 1 | 74 |
| 944 | Phi-3 Mini 4K Instruct | Microsoft | 1449 ± 1 | 118 |
| 984 | Phi-4 Multimodal Instruct | Microsoft | 1440 ± 1 | 138 |
| 1003 | Phi-3-mini-4K-instruct-cpo-simpo | Microsoft | 1436 ± 1 | 130 |
| 1049 | Phi-3.5-mini-instruct | Microsoft | 1424 ± 1 | 136 |
| 1106 | Phi-3 Mini 128K Instruct | Microsoft | 1406 ± 1 | 98 |
Interactive version: theaggregate.ai/company?slug=microsoft · How It Works · Data refreshed daily, snapshot 2026-09-05.