ByteDanceSeed-Coder
Seed-Coder 8B Base
Seed-Coder 8B Base supports code completion and fill-in-the-middle generation.
Model design
The base checkpoint was trained on six trillion tokens from GitHub and code-related web material. The project uses language models to help filter its training data. Its context limit is 32,768 tokens.Seed-Coder 8B Base official model card
Practical use
Practical guidance. For an editor, pass both the code before the cursor and the code after it through the documented infill format. A plain chat request does not reproduce that task.
Specifications & access
| Detail | Value and conditions |
|---|---|
| Access | Downloadable |
| Model identifiers | Seed-Coder-8B-Base |
| Context window | 32,768 tokens Published checkpoint context limit. Seed-Coder 8B Base official model card |
| License | MITSeed-Coder 8B Base official model card |
| Access | Weights are published in the official ByteDance Seed repository.Seed-Coder 8B Base official model card |
Related models
- Seed-Coder 8B Instruct
ByteDance · Seed-Coder
Seed-Coder 8B Instruct adapts the base coding model to follow user requests.
Launch date not confirmedDownloadable - Stable-DiffCoder 8B Base
ByteDance · Stable-DiffCoder
Stable-DiffCoder 8B Base is a masked-diffusion coding model derived from Seed-Coder.
Launch date not confirmedDownloadable