Poro 34B

Poro 2 Long Instruct is an instruction-following chatbot model with extended context support, created through supervised fine-tuning (SFT) of the Poro 2 Long Base model followed by merging the SFT checkpoint back with the base model to preserve long-context performance. This model is designed for conversational AI applications and instruction following in both Finnish and English, with support for context lengths up to 128K tokens. It was trained on a carefully curated mix of English and Finnish instruction data.

Poro 2 8B Math Reasoning RL Preview is a specialized model focused on mathematical reasoning and problem-solving. This preview model was created through reinforcement learning (RL) on top of the Math Reasoning SFT checkpoint. This model excels at mathematical reasoning tasks but is not optimized for general conversational use or other domains.

Poro 2 Long Base is an 8B parameter decoder-only transformer created by extending Poro 2 8B Base from an 8K to 128K token context window using LongRoPE. The model supports both English and Finnish with an extended context window of 128K tokens. Poro 2 Long Base is released as a fully open source model under the Llama 3.1 Community License.

Poro 2 8B Math Reasoning SFT Preview is a specialized model focused on mathematical reasoning and problem-solving. This preview model was created through supervised fine-tuning of a context-extended base model. This model excels at mathematical reasoning tasks but is not optimized for general conversational use or other domains.

Note for most users: This is an intermediate checkpoint from our post-training pipeline. Most users should use Poro 2 70B Instruct instead, which includes an additional round of Direct Preference Optimization (DPO) for improved response quality and alignment. This SFT-only model is primarily intended for researchers interested in studying the effects of different post-training techniques.

Note for most users: This is an intermediate checkpoint from our post-training pipeline. Most users should use Poro 2 8B Instruct instead, which includes an additional round of Direct Preference Optimization (DPO) for improved response quality and alignment. This SFT-only model is primarily intended for researchers interested in studying the effects of different post-training techniques.