Phi-3.5-Vision Instruct: benchmark results
Provider: Microsoft. Released 2024-08-20. Access: Open.
Aggregate Rating 1397 ± 15, rank #6250 of 10916 rated models, from 207 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| Epoch AI - Science Qa | 91.3 | Score | 100 |
| MiSCHiEF (Culture) - Dual Caption-Image Alignment | 57.71 | Accuracy (%, 0-100) on MiC: 279 culture minimal pairs, the s | 100 |
| MultivationBench - Reiss Practical Motivation | 28.9 | Exact-match accuracy (%) of multi-label motivation predictio | 85.7 |
| ThermEval - Colorbar Position | 100 | Accuracy (times 100, 0-100) on Task 4, the colorbar's side ( | 83.3 |
| ThermEval - Temperature Estimation (Coordinates) | 3.14 | Mean absolute error (degrees Celsius, 0 or more) on Task 6, | 78.3 |
| MERA Multi - ruNaturalScienceVQA | 42.6 | Score (%) | 78 |
| MERA Multi - WEIRD | 50.5 | Score (%) | 76.3 |
| DatBench - Document Understanding | 60.7 | Accuracy (%) | 76 |
| MiSCHiEF (Culture) - Caption-to-Image Matching | 82.8 | Accuracy (%, 0-100) on MiC: 279 culture minimal pairs, the s | 75 |
| ThermEval - Colorbar Detection | 100 | Accuracy (times 100, 0-100) on Task 4, whether a ThermEval-D | 75 |
| GeoMMBench - Spatial | 54.1 | Micro-averaged accuracy (%) on the spatial relation question | 73.5 |
| MERA Multi - LabTabVQA | 15 | Score (%) | 72.9 |
Interactive version: theaggregate.ai/model?slug=phi-3-5-vision-instruct · How It Works · Data refreshed daily, snapshot 2026-11-08.