Llama 3 Youko 8B
Youko 8B continues pretraining Meta Llama 3 8B on 22B Japanese and English tokens.
Research release
A bilingual training stage follows Meta's base
The model retains 32 layers with hidden width 4,096. rinna's contribution is continued language training. The name Youko refers to a fox-like creature in Japanese folklore.Llama 3 Youko 8B official model card
Use the base for language adaptation studies
Practical guidance. Compare Japanese, English and mixed-language continuations before applying task-specific fine-tuning. That comparison helps distinguish changes in language fluency from changes caused by later instruction training.
Specifications & access
| Detail | Value and conditions |
|---|---|
| Status | Research release |
| Access | Downloadable |
| Model identifiers | llama-3-youko-8bllama-3-youko-8b-gptq |
| Ancestry | meta-llama/Meta-Llama-3-8B.Llama 3 Youko 8B official model card |
| License | Meta Llama 3 Community License.Llama 3 Youko 8B official model card |
Related models
- Llama 3 Youko 70B
rinna · Youko
Youko 70B continues pretraining Meta Llama 3 70B on 5B Japanese and English tokens.
Launch date not confirmedResearch releaseDownloadable - Llama 3 Youko 8B Instruct
rinna · Youko
Youko 8B Instruct combines supervised fine-tuning with a chat vector from Meta's Llama 3 instruction model, followed by DPO.
Launch date not confirmedDownloadable