Llama 3 Youko 70B Instruct
Youko 70B Instruct combines supervised fine-tuning with a chat vector from Meta's Llama 3 instruction model.
A chat vector transfers instruction behavior
The published merge adds half of the difference between Llama 3 70B Instruct and its base to rinna's SFT model. It excludes the embedding layer. The card describes SFT and merging; it does not identify a DPO stage for this size.Llama 3 Youko 70B Instruct official model card
Do not copy the 8B training recipe into this record
Practical guidance. The two Youko instruction sizes have different documented training stages. If the deployment depends on a specific alignment method, verify that size's card instead of assuming the whole family shares it.
Specifications & access
| Detail | Value and conditions |
|---|---|
| Access | Downloadable |
| Model identifiers | llama-3-youko-70b-instructllama-3-youko-70b-instruct-gptq |
| License | Meta Llama 3 Community License.Llama 3 Youko 70B Instruct official model card |
| Prompt format | Llama 3 chat format.Llama 3 Youko 70B Instruct 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