Poro 34B OpenAssistant research release
This experimental Poro adaptation studies bilingual instruction tuning with OpenAssistant data. The card recommends Poro 34B Chat for normal conversation.
Research release
Model design and training
The supervised data combines curated OpenAssistant 2 conversations with Finnish translations made by Poro. The card also publishes a HelpSteer2 preference-training recipe, using helpfulness scores to form pairs. It describes both SFT and DPO experiments; the page does not establish a separate public model for every recipe.Poro-34B-chat-OpenAssistant: official model card
Practical assessment
Practical guidance. This record is useful for reproducing an instruction-data experiment. Keep the checkpoint revision and the exact training recipe separate in your notes. A recipe printed beside a checkpoint does not prove that every listed training stage is present in that file.
Specifications & access
| Detail | Value and conditions |
|---|---|
| Status | Research release |
| Access | Downloadable |
| Model identifiers | Poro-34B-chat-OpenAssistant |
| License | Apache 2.0Poro-34B-chat-OpenAssistant: official model card |
| Parent | Poro 34B.Poro-34B-chat-OpenAssistant: official model card |
| Research focus | English and Finnish instruction data.Poro-34B-chat-OpenAssistant: official model card |
Related models
- Poro 34B Chat
LumiOpen ยท Poro
A Finnish and English instruction model adapted from Poro 34B with supervised fine-tuning and ChatML formatting.
Launch date not confirmedDownloadable