Doubao-Seed-1.6 (Thinking): benchmark results
Provider: ByteDance. Released 2025-06-11. Access: API.
Aggregate Rating 1626 ± 13, rank #621 of 2170 rated models, from 99 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| AEGIS (Academic Image Forensics) | 46.66 | Normalized Forensic Index (%): harmonic mean of the macro-F1 | 92 |
| AEGIS (Academic Image Forensics) - Manipulation Classification | 41.85 | Macro-F1 (%) over three classes, naming a highlighted edited | 92 |
| Embodied3DBench - Spatial Relation (3D) | 72.2 | Spatial Relation (3D) task (object-to-object and object-to-c | 91.7 |
| VehicleMemBench (Gold Memory) | 85.4 | Exact state match (%): share of VehicleMemBench's 500 execut | 90.9 |
| EmoHarbor | 4.02 | Mean of 10 Rubric Dimensions (1-5 Likert) | 89.5 |
| AGI-Eval Community - Cognition | 83.98 | Accuracy (%) | 89.3 |
| AGI-Eval Community - Cognition (English) | 85.56 | Accuracy (%) | 88 |
| ScholScan (Image Input) - Computation and Formulae | 10.2 | Flaw-detection score (0-1, scaled to 0-100): per question, z | 87.5 |
| AGI-Eval Community - Cognition (Chinese) | 82.84 | Accuracy (%) | 86.2 |
| ScholScan (Text Input) - Language and Expression | 4.2 | Flaw-detection score (0-1, scaled to 0-100): per question, z | 85.7 |
| SuperCLUE-VLM (June 2025) - Overall | 65.67 | Score | 84.6 |
| PHYBench | 45.53 | EED Score | 80.6 |
Interactive version: theaggregate.ai/model?slug=doubao-seed-1-6-thinking · How It Works · Data refreshed daily, snapshot 2026-11-05.