HyperCLOVAX-SEED-Vision-Instruct-3B: benchmark results
Provider: Naver. Access: API.
Aggregate Rating 1434 ± 25, rank #4500 of 10916 rated models, from 38 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| KOFFVQA - Hallucination and Robustness | 80 | Score (%) | 80.9 |
| KOFFVQA - Relationship | 68.67 | Score (%) | 74.1 |
| KOFFVQA - Korean OCR | 75 | Score (%) | 61.7 |
| KOFFVQA - Object Attributes | 73.83 | Score (%) | 61.1 |
| KOFFVQA - Document Understanding | 56.33 | Score (%) | 51.2 |
| HAERAE-Vision - Natural Objects | 16.21 | Checklist Score (%; original under-specified queries, GPT-5- | 45.5 |
| KOFFVQA - Korean Recognition | 20 | Score (%) | 44.4 |
| KOFFVQA | 59.6 | Overall Score (%) | 43.2 |
| HAERAE-Vision - Health/Medical | 29.74 | Checklist Score (%; original under-specified queries, GPT-5- | 36.4 |
| KOFFVQA - Table Understanding | 35.33 | Score (%) | 35.8 |
| KOFFVQA - Commonsense Reasoning | 67.56 | Score (%) | 33.3 |
| KOFFVQA - Recognition | 50 | Score (%) | 29.6 |
Interactive version: theaggregate.ai/model?slug=hyperclovax-seed-vision-instruct-3b · How It Works · Data refreshed daily, snapshot 2026-11-08.