Model Card 2026 Technical report

Seed1.8 Model Card: Towards Generalized Real-World Agency

A team model card describing a foundation model for multi-turn interaction, tool use, and complex execution.

Bytedance Seed

The formal author is Bytedance Seed, and Qinyu Luo is named in the Contributions section. The model was released in December 2025; this entry uses the report’s March 2026 arXiv publication year.

Seed1.8 evaluation framework mapping real user needs to general agentic search, tool use, coding, reasoning, writing, vision, and internal benchmarks. — Fig. 12 — Seed1.8 evaluation system: mapping real user needs to benchmark categories.Original source ↗

01 / The question

What does a general-purpose foundation model need to complete real-world workflows?

02 / The approach

Seed1.8

The report presents a unified agentic interface spanning search, code, and GUI interaction, together with configurable inference modes. Its evaluations cover language, vision, agentic capabilities, efficiency, and application-oriented workflows.

Read the full paper

03 / Citation

Cite this work

BibTeX

@misc{seed2026seed18,
  title = {Seed1.8 Model Card: Towards Generalized Real-World Agency},
  author = {{Bytedance Seed}},
  year = {2026},
  eprint = {2603.20633},
  archivePrefix = {arXiv},
  primaryClass = {cs.AI},
  url = {https://arxiv.org/abs/2603.20633}
}

Select the citation to copy it manually.

04 / Versions & records

Publication record

1 Google Scholar record

Seed1.8 Model Card: Towards Generalized Real-World Agency ↗

READ HERE · 51 PAGES

Full paper

Document

Loading document…

Scroll through every page. Select text to copy, or use search to find a passage.

More research

View all

What does a general-purpose foundation model need to complete real-world workflows?

Seed1.8

Cite this work

Publication record

Full paper

More research

Seed2.0↗

Seed2.1↗

Seed2.0

Seed2.1