InternVL3-2B-Instruct: benchmark results
Provider: Shanghai AI Lab. Access: Open.
Aggregate Rating 1395 ± 34, rank #6311 of 10916 rated models, from 36 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| DatBench - Spatial Reasoning | 22 | Accuracy (%) | 80.8 |
| DatBench - Scene OCR | 54.5 | Accuracy (%) | 65.4 |
| DatBench - General VQA | 42.5 | Accuracy (%) | 50 |
| DatBench - Counting | 89.5 | Accuracy (%) | 38.5 |
| FewMMBench (8-Shot, Random) | 67.3 | Mean accuracy (%) on the FewMMBench query sets (about 250 mu | 36 |
| DatBench - Diagrams & Tables | 24 | Accuracy (%) | 34.6 |
| FewMMBench (8-Shot, Similar) | 66.6 | Mean accuracy (%) on the FewMMBench query sets (about 250 mu | 34 |
| DatBench - Document Understanding | 25.7 | Accuracy (%) | 32 |
| FewMMBench (4-Shot, Similar + CoT) | 57.7 | Mean accuracy (%) on the FewMMBench query sets (about 250 mu | 32 |
| FewMMBench (4-Shot, Similar) | 66.7 | Mean accuracy (%) on the FewMMBench query sets (about 250 mu | 32 |
| FewMMBench (8-Shot, Similar + CoT) | 56.8 | Mean accuracy (%) on the FewMMBench query sets (about 250 mu | 32 |
| DatBench - Math & Logic | 12.1 | Accuracy (%) | 30.8 |
Interactive version: theaggregate.ai/model?slug=internvl3-2b-instruct · How It Works · Data refreshed daily, snapshot 2026-11-08.