InternVL3.5-14B: benchmark results
Provider: Shanghai AI Lab. Released 2025-08-26. Access: Open.
Aggregate Rating 1561 ± 8, rank #1530 of 10916 rated models, from 454 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| AgroTools - Summary Accuracy | 33.96 | Summary accuracy (0-100): normalized similarity of the final | 100 |
| HM-Bench (PCA) | 43.08 | Average accuracy (%) over all tasks of HM-Bench's 19,337 hyp | 100 |
| HM-Bench (Report) | 39.52 | Average accuracy (%) over all tasks of HM-Bench's 19,337 hyp | 100 |
| HM-Bench (VSR2) | 45.09 | Average accuracy (%) over all tasks of HM-Bench's 19,337 hyp | 100 |
| MVX-Bench - Re-Identification | 62 | Accuracy (%) on the 150 Re-Identification questions (decide | 100 |
| TRAP - Text Task Accuracy | 91 | Task accuracy (%) on the 100 text-document scenarios of 500 | 100 |
| EgoCoT-Bench - Hand-Object Temporal Retrospection | 75 | Answer accuracy (%, four-way multiple choice, strict exact m | 97.2 |
| SpatialUAV - Global Motion | 55.3 | Motion Description Similarity (%; GPT-5.4-mini judge, 0-1 pe | 94.1 |
| HM-Bench (RGB) | 42.88 | Average accuracy (%) over all tasks of HM-Bench's 19,337 hyp | 93.8 |
| PanoSpace-Bench - Observer Distance | 56.8 | Accuracy (%; 250 three- or four-option multiple-choice quest | 90.9 |
| DistortBench - Color | 56.4 | Joint accuracy (%) on the five color distortions (diffusion, | 88.2 |
| EPIC-Bench - Spatial-Related Attributes | 49.96 | Target localization score (0-100) on the spatial-attribute t | 87.5 |
Interactive version: theaggregate.ai/model?slug=internvl3-5-14b · How It Works · Data refreshed daily, snapshot 2026-11-08.