LLM-jp / NII LLMCLLM-jp-3
LLM-jp-3 980M Base
LLM-jp-3 980M Base is the 980M pretrained text model in LLM-jp's generation-3 size series. The underlying model was exposed to 2.1 trillion pretraining tokens.
Research release
The size changes the architecture
The 20-layer model has hidden width 1,536 and eight attention heads. Pretraining includes Japanese web and archive material, English Dolma sources and code from The Stack.LLM-jp-3 980M Base official model card
Use the size as an experimental control
Practical guidance. Use this checkpoint to examine behavior below the billion-parameter scale. Measure memory and accuracy on the same workload as the 440M model.
Specifications & access
| Detail | Value and conditions |
|---|---|
| Status | Research release |
| Access | Downloadable |
| Model identifiers | llm-jp-3-980m |
| Context window | 4,096 tokens Maximum sequence length stated in the model card. LLM-jp-3 980M Base official model card |
| Architecture | 20 layers; hidden size 1,536; 8 attention heads.LLM-jp-3 980M Base official model card |
| Parameters | 989,882,880 total, from the card's embedding and non-embedding counts.LLM-jp-3 980M Base official model card |
| License | Apache 2.0; research weights.LLM-jp-3 980M Base official model card |
Related models
- LLM-jp-3 980M Instruct3
LLM-jp / NII LLMC ยท LLM-jp-3
LLM-jp-3 980M Instruct3 is a preference-aligned instruction release with 4,096-token context and Apache 2.0 weights.
Launch date not confirmedResearch releaseDownloadable