OpenVLThinker-7B: benchmark results
Provider: Other.
Aggregate Rating 1511 ± 38, rank #2181 of 10916 rated models, from 15 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| VS-Bench (VisualSwap) - MathVista | 42.5 | Probe accuracy (%; 200 image pairs sourced from MathVista: a | 71.4 |
| VS-Bench (VisualSwap) - MathVerse | 35 | Probe accuracy (%; 200 image pairs sourced from MathVerse: a | 64.3 |
| HID-Chart - Math | 39.2 | Accuracy (%; GPT-5 mini judge, 97 math questions) | 46.7 |
| HID-Chart - Biology | 36 | Accuracy (%; GPT-5 mini judge, 136 biology questions) | 43.3 |
| HID-Chart - Physics | 38.7 | Accuracy (%; GPT-5 mini judge, 119 physics questions) | 43.3 |
| HID-Chart - Social | 37.9 | Accuracy (%; GPT-5 mini judge, 335 social questions) | 40 |
| VS-Bench (VisualSwap) | 28.6 | Probe accuracy (%; mean over 800 image pairs, 200 from each | 35.7 |
| HID-Chart | 38.5 | Accuracy (%; GPT-5 mini judge against the annotated answer, | 31.6 |
| MathVision | 25.9 | Overall Accuracy (%) | 27.4 |
| HID-Chart - Computer Science | 41.2 | Accuracy (%; GPT-5 mini judge, 323 computer science question | 26.7 |
| HID-Chart - Engineering | 35.3 | Accuracy (%; GPT-5 mini judge, 116 engineering questions) | 26.7 |
| HID-Chart - Finance | 38.2 | Accuracy (%; GPT-5 mini judge, 327 finance questions) | 23.3 |
Interactive version: theaggregate.ai/model?slug=openvlthinker-7b · How It Works · Data refreshed daily, snapshot 2026-11-08.