InternVL3.5-2B: benchmark results
Provider: Shanghai AI Lab. Released 2025-08-26. Access: Open.
Aggregate Rating 1469 ± 11, rank #2869 of 10916 rated models, from 334 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| Jina VLM Release - LogicVista | 41.4 | Accuracy (%) | 100 |
| Jina VLM Release - MathVerse Vision Only | 35.3 | Accuracy (%) | 100 |
| Jina VLM Release - MathVision | 26.5 | Accuracy (%) | 100 |
| Jina VLM Release - MathVista | 61.5 | Accuracy (%) | 100 |
| Jina VLM Release - WeMath | 19.1 | Accuracy (%) | 100 |
| UXBench (Mobile) - Content Mismatch | 72 | Accuracy (0-100) on service names inconsistent with the page | 100 |
| EgoCoT-Bench - Hand-Object Association | 63.16 | Answer accuracy (%, four-way multiple choice, strict exact m | 97.2 |
| TRAP - Text Privacy | 19 | Privacy score (%, 1 minus leakage) on the 100 text-document | 95.2 |
| UXBench (Mobile) - Function Mismatch | 76 | Accuracy (0-100) on descriptions inconsistent with the provi | 90.9 |
| EgoCoT-Bench - Hand-Object Temporal Retrospection | 71.88 | Answer accuracy (%, four-way multiple choice, strict exact m | 88.9 |
| EgoCoT-Bench - Location and Visibility Retrospection | 66.94 | Answer accuracy (%, four-way multiple choice, strict exact m | 88.9 |
| SYNCR - Spatial Measurement | 46 | Exact-match accuracy (%) on the SYNCR Spatial Measurement ta | 88.2 |
Interactive version: theaggregate.ai/model?slug=internvl3-5-2b · How It Works · Data refreshed daily, snapshot 2026-11-08.