Polyglot-Ko 3.8B
A Korean autoregressive language model developed by the EleutherAI Polyglot team. The released weights support text completion and further adaptation.
Research release
Korean modeling at this scale
Polyglot-Ko 3.8B increases depth to 32 layers and hidden width to 3,072. Its architecture uses rotary positions and a Korean vocabulary of 30,003 tokens, padded to 30,080 embedding entries.polyglot-ko-3.8b model cardpolyglot-ko-3.8b model configuration
Training data and use
The release reports Korean benchmark results with different numbers of examples in each prompt. Those results depend on the prompt setup. They should not be treated as scores from an instruction-tuned chat model.polyglot-ko-3.8b model card
Specifications & access
| Detail | Value and conditions |
|---|---|
| Status | Research release |
| Access | Downloadable |
| Model identifiers | EleutherAI/polyglot-ko-3.8b |
| Context window | 2,048 tokens Maximum position count in the released model configuration. polyglot-ko-3.8b model configuration |
| Parameters | 3,809,974,272polyglot-ko-3.8b model card |
| Language | Koreanpolyglot-ko-3.8b model card |
| License | Apache 2.0polyglot-ko-3.8b model card |
Related models
- Polyglot-Ko 1.3B
EleutherAI · Polyglot-Ko
A compact Korean base model from the Polyglot project.
Launch date not confirmedResearch releaseDownloadable - Polyglot-Ko 12.8B
EleutherAI · Polyglot-Ko
The largest documented Polyglot-Ko base model, trained on Korean text.
Launch date not confirmedResearch releaseDownloadable - Polyglot-Ko 5.8B
EleutherAI · Polyglot-Ko
A wider Korean base model with 28 layers and downloadable weights.
Launch date not confirmedResearch releaseDownloadable