Granite Vision 4.1 4B: benchmark results
Provider: IBM. Released 2026-04-29. Access: Open.
Aggregate Rating 1491 ± 52, rank #2433 of 10552 rated models, from 12 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| ParseBench - Charts | 47.47 | ChartDataPointMatch Score | 62 |
| VAREX-Bench | 94.2 | Exact Match (imported) | 56.2 |
| PhysCount - Image Formation NAE | 41.51 | Image-formation normalized absolute error (%) | 20.5 |
| PhysCount - Appearance NAE | 34.73 | Appearance normalized absolute error (%) | 17.9 |
| PhysCount - Entity NAE | 45.38 | Entity normalized absolute error (%) | 15.4 |
| PhysCount - Group-Acc | 0.61 | Group accuracy (%): groups with every answer exact | 11.5 |
| PhysCount - Score | 9.02 | Score (%): group-macro product of one minus NAE over questio | 10.3 |
| ParseBench | 39.45 | Overall Score | 7.4 |
| ParseBench - Tables | 63.81 | GTRM Score (GriTS + TableRecordMatch) | 6.6 |
| PhysCount - Correspondence NAE | 67.96 | Correspondence normalized absolute error (%) | 5.1 |
| ParseBench - Semantic Formatting | 21.52 | Semantic Formatting Score | 4.1 |
| ParseBench - Content Faithfulness | 64.43 | Content Faithfulness Score | 0 |
Interactive version: theaggregate.ai/model?slug=granite-vision-4-1-4b · How It Works · Data refreshed daily, snapshot 2026-11-16.