LLM-jp-3 3.7B Instruct2
LLM-jp-3 3.7B Instruct2 is the 3.7B supervised instruction release in LLM-jp's generation-3 size series. The underlying model was exposed to 2.1 trillion pretraining tokens.
Research release
Supervised alignment before DPO
Instruct2 applies supervised fine-tuning. Its data includes ichikara, AnswerCarefully, Magpie, Daring-Anteater, FLAN and Japanese-English code instructions. It is the SFT comparison point for Instruct3.LLM-jp-3 3.7B Instruct2 official model card
Compare before and after preference training
Practical guidance. Pair this 3.7B checkpoint with Instruct3 for a same-size alignment comparison. Check task completion, refusals and formatting separately.
Specifications & access
| Detail | Value and conditions |
|---|---|
| Status | Research release |
| Access | Downloadable |
| Model identifiers | llm-jp-3-3.7b-instruct2 |
| Context window | 4,096 tokens Maximum sequence length stated in the model card. LLM-jp-3 3.7B Instruct2 official model card |
| Architecture | 28 layers; hidden size 3,072; 24 attention heads.LLM-jp-3 3.7B Instruct2 official model card |
| Parameters | 3,782,317,056 total, from the card's embedding and non-embedding counts.LLM-jp-3 3.7B Instruct2 official model card |
| License | Apache 2.0; research weights.LLM-jp-3 3.7B Instruct2 official model card |
Related models
- LLM-jp-3 3.7B Base
LLM-jp / NII LLMC · LLM-jp-3
LLM-jp-3 3.7B Base is a pretrained text model with 4,096-token context and Apache 2.0 weights.
Launch date not confirmedResearch releaseDownloadable - LLM-jp-3 3.7B Instruct3
LLM-jp / NII LLMC · LLM-jp-3
LLM-jp-3 3.7B Instruct3 is a preference-aligned instruction release with 4,096-token context and Apache 2.0 weights.
Launch date not confirmedResearch releaseDownloadable