Doubao-Seed-1.6-vision: benchmark results
Provider: ByteDance.
Aggregate Rating 1699 ± 17, rank #353 of 10916 rated models, from 87 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| Dr. DocBench - Formula CDM | 41.28 | Character Detection Matching score of display formulas (0-10 | 100 |
| FlagEval EmbodiedVerse - BLINK (val) | 79.91 | Score | 100 |
| FlagEval EmbodiedVerse - CV-Bench (test) | 89.81 | Score | 100 |
| FlagEval EmbodiedVerse - RealWorldQA | 75.42 | Score | 97.8 |
| FlagEval EmbodiedVerse - RoboSpatial-Home | 63.95 | Score | 95.7 |
| Dr. DocBench - Formula Edit Distance | 31 | Normalized edit distance of display formulas (normalized LaT | 95.5 |
| SIS-Bench - Action Prediction | 65.8 | Accuracy (%; 263 action prediction questions; four-option mu | 94 |
| FlagEval VQA - Text Understanding (Chinese) | 78.6 | Score | 93.5 |
| EPIC-Bench - Contact Relationship | 47.48 | Manipulation score (0-100) on contact relationship (localizi | 93.2 |
| EPIC-Bench | 46.8 | Average score (0-100) over all samples, each sample scored b | 92 |
| EPIC-Bench - Basic Attributes | 54.64 | Target localization score (0-100) on the basic-attribute tas | 92 |
| EPIC-Bench - Embodied Compositional Attributes | 42.27 | Target localization score (0-100) on the embodied compositio | 92 |
Interactive version: theaggregate.ai/model?slug=doubao-seed-1-6-vision · How It Works · Data refreshed daily, snapshot 2026-11-08.