Seed 1.8: benchmark results
ByteDance Seed 1.8 (Doubao) general-purpose model. Provider: ByteDance. Released 2025-12-18. Access: API.
Aggregate Rating 1733 ± 6, rank #139 of 1656 rated models, from 343 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| ALPBench - Badminton Rackets (Attribute F1) | 65.92 | Attribute-level F1 (%): the F1 of each target attribute comp | 100 |
| ALPBench - Badminton Rackets (Profile F1) | 25.5 | Profile-level F1 (%): the harmonic mean of the shares of pre | 100 |
| ALPBench - Baijiu (Attribute F1) | 65.12 | Attribute-level F1 (%): the F1 of each target attribute comp | 100 |
| ALPBench - Baijiu (Profile F1) | 20.08 | Profile-level F1 (%): the harmonic mean of the shares of pre | 100 |
| ALPBench - Pants (Attribute F1) | 56.56 | Attribute-level F1 (%): the F1 of each target attribute comp | 100 |
| ALPBench - Snacks (Attribute F1) | 31.1 | Attribute-level F1 (%): the F1 of each target attribute comp | 100 |
| BrowseComp-V3 - Culture | 38.67 | Success rate (%): share of questions answered correctly on t | 100 |
| BrowseComp-V3 - Process Score | 58.44 | Process score (%): mean share of each question's expert-anno | 100 |
| BrowseComp-V3 - Society | 37.33 | Success rate (%): share of questions answered correctly on t | 100 |
| Chronicles-OCR - Archaic Script Classification | 96.7 | Accuracy (%) of classifying the image into one of the Seven | 100 |
| CourtSI-Bench - Object-Line Distance | 43.8 | Threshold mean relative accuracy (%), averaged over confiden | 100 |
| FinReasoning (Deep Insight) - Architectural Richness and Strictness | 93.5 | Rubric score (0-100): a judge score fused from DeepSeek-V3 a | 100 |
Interactive version: theaggregate.ai/model?slug=seed-1-8 · How It Works · Data refreshed daily, snapshot 2026-10-20.