DeepEyesV2: benchmark results
Provider: Other.
Aggregate Rating 1562 ± 43, rank #1522 of 10916 rated models, from 32 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| VTC-Bench (Code Tools) - Measurement | 44.76 | Pass rate (%) on measurement (105 questions on size, positio | 83.3 |
| VTC-Bench (Code Tools) - Spatial Reasoning | 44.44 | Pass rate (%) on spatial reasoning under extreme conditions | 58.3 |
| AdaptMMBench - Real-world | 56.33 | Accuracy (%) on the Real-world domain of AdaptMMBench, enhan | 55 |
| AdaptMMBench (Text Reasoning) | 52.08 | Overall accuracy (%) on the 1,300 AdaptMMBench samples (auxi | 50 |
| VTC-Bench (Code Tools) - Math | 28.18 | Pass rate (%) on geometry math with auxiliary constructions | 44.4 |
| AdaptMMBench - GUI | 55.67 | Accuracy (%) on the GUI domain of AdaptMMBench, enhancement- | 40 |
| VTC-Bench (Interface Tools) - Perceptual Restoration | 28 | Pass rate (%) on perceptual restoration (50 questions under | 38.9 |
| VTC-Bench (Code Tools) | 30 | Average pass rate (%) answering VTC-Bench's 680 visual quest | 33.3 |
| VTC-Bench (Code Tools) - Robust OCR | 20 | Pass rate (%) on robust OCR (50 questions on degraded text), | 33.3 |
| AdaptMMBench (Oracle Visual Evidence) | 70.23 | Overall accuracy (%) on the 1,300 AdaptMMBench samples (auxi | 30 |
| AdaptMMBench - Knowledge | 53.5 | Accuracy (%) on the Knowledge domain of AdaptMMBench, enhanc | 30 |
| VTC-Bench (Code Tools) - Chart | 30 | Pass rate (%) on chart reasoning (100 questions), answering | 25 |
Interactive version: theaggregate.ai/model?slug=deepeyesv2 · How It Works · Data refreshed daily, snapshot 2026-11-08.