LLM-jp-3 172B Instruct2
LLM-jp-3 172B Instruct2 is the supervised fine-tuning release at the largest dense generation-3 size. It retains the 2.1-trillion-token pretraining foundation.
Restricted access
The SFT release has its own access conditions
The public model page requires agreement to the access conditions and sharing Hugging Face profile information. Its SFT mixture includes Japanese instruction and safety data, synthetic conversation, English instruction data and code examples.LLM-jp-3 172B Instruct2 official model card
Do not transfer Instruct3 terms to Instruct2
Practical guidance. Use this model when a same-size SFT comparison is required. Check its own license and download gate rather than assuming that the newer Instruct3 permissions apply.
Specifications & access
| Detail | Value and conditions |
|---|---|
| Status | Restricted access |
| Access | Downloadable |
| Model identifiers | llm-jp-3-172b-instruct2 |
| Context window | 4,096 tokens Maximum sequence length stated in the model card. LLM-jp-3 172B Instruct2 official model card |
| Architecture | 96 layers; hidden size 12,288; 96 attention heads.LLM-jp-3 172B Instruct2 official model card |
| License | Custom llm-jp-3-172b-instruct2 terms; gated download.LLM-jp-3 172B Instruct2 official model card |
| Training stage | Supervised fine-tuning, before the Instruct3 DPO stage.LLM-jp-3 172B Instruct2 official model card |
Related models
- LLM-jp-3 172B Base
LLM-jp / NII LLMC · LLM-jp-3 172B
The final 2.1-trillion-token 172B base checkpoint uses application-based access and custom terms.
Launch date not confirmedRestricted accessDownloadable - LLM-jp-3 172B Instruct3
LLM-jp / NII LLMC · LLM-jp-3 172B
The 172B preference-aligned release has a separate license that permits redistribution subject to attribution and other terms.
Launch date not confirmedResearch releaseDownloadable