DreadPoor/Paxinium-12b-Model_Stock

Hugging Face
TEXT GENERATIONConcurrent Unit Cost:1Model Size:12BQuant:FP8Context Size:32kTool Calling:SupportedPublished:May 5, 2025Architecture:Transformer0.0K Featherless Exclusive Warm

DreadPoor/Paxinium-12b-Model_Stock is a 12 billion parameter language model created by DreadPoor, formed by merging several pre-trained models using the Model Stock method. It utilizes Delta-Vector/Francois-PE-V2-Huali-12B as its base model, integrating capabilities from Rei-V3-KTO-12B, EtherealAurora-12B-v2, GodSlayer-12B-ABYSS, and BlueLight-12B. This model is designed for general language tasks, leveraging the combined strengths of its constituent models.

Loading preview...

DreadPoor/Paxinium-12b-Model_Stock Overview

DreadPoor/Paxinium-12b-Model_Stock is a 12 billion parameter language model developed by DreadPoor, created through a sophisticated merging process. This model was constructed using the Model Stock merge method, a technique designed to combine the strengths of multiple pre-trained language models into a single, more capable entity.

Merge Details

The foundation of Paxinium-12b-Model_Stock is Delta-Vector/Francois-PE-V2-Huali-12B. This base model was then enhanced by integrating the following distinct models:

  • Delta-Vector/Rei-V3-KTO-12B
  • yamatazen/EtherealAurora-12B-v2
  • redrix/GodSlayer-12B-ABYSS
  • yamatazen/BlueLight-12B

This merging strategy aims to consolidate diverse linguistic knowledge and capabilities from its constituent models, resulting in a versatile language model. The merging process was configured with normalize: true for parameters and utilized bfloat16 data type.

Intended Use Cases

Given its composition from multiple general-purpose language models, Paxinium-12b-Model_Stock is suitable for a broad range of natural language processing tasks. Developers can leverage its combined knowledge for applications requiring robust language understanding and generation.