Poro 2 8B Base
Poro 2 8B Base continues pretraining Meta's Llama 3.1 8B to strengthen Finnish while retaining English, code and mathematics. It is a trained derivative, not unchanged Llama hosting.
Model design and training
Continued pretraining used 165 billion tokens. The mix allocates roughly 30% each to Finnish, English and code, with 10% mathematics. Finnish material comes from FineWeb2; the other components include FineWeb-Edu, StarCoder data and FineMath. The release precedes instruction and preference tuning.Llama-Poro-2-8B-base: official model card
Practical assessment
Practical guidance. This base suits controlled Finnish adaptation when model size matters. Compare it with the original Llama checkpoint on both Finnish and English tasks to see what continued pretraining changes. Use the Instruct release when conversation behavior is required.
Specifications & access
| Detail | Value and conditions |
|---|---|
| Access | Downloadable |
| Model identifiers | Llama-Poro-2-8B-base |
| Context window | 8,192 tokens Sequence length reported by this release's model card; this is not a guaranteed useful retrieval length. Llama-Poro-2-8B-base: official model card |
| Model size | 8.03 billion parameters; 32 decoder layers.Llama-Poro-2-8B-base: official model card |
| License | Llama 3.1 Community License.Llama-Poro-2-8B-base: official model card |
Related models
- Poro 2 8B Instruct
LumiOpen · Poro 2
An 8B Finnish and English conversation model trained with supervised instructions and Direct Preference Optimization.
Launch date not confirmedDownloadable - Poro 2 8B SFT
LumiOpen · Poro 2
An intermediate instruction-tuned Poro 2 8B checkpoint before the final preference-training stage.
Launch date not confirmedResearch releaseDownloadable