Naphula/Goetia-26B-A4B-v1.5
Naphula/Goetia-26B-A4B-v1.5 is a 26 billion parameter language model based on the Gemma-4 architecture, created by Naphula through a merge of multiple pre-trained models using the moe_della method. This model integrates diverse capabilities from its constituent models, including instruction-tuned and specialized adaptations, to offer a broad range of generative text functionalities. It is designed for general-purpose text generation with a 32768 token context length, suitable for various applications requiring nuanced language understanding and generation.
Loading preview...
Overview
Naphula/Goetia-26B-A4B-v1.5 is a 26 billion parameter language model built upon the Gemma-4-26B-A4B-it base model. It was created by Naphula using the moe_della merge method from mergekit-exp, combining over 25 distinct pre-trained models and adapters. This extensive merging process aims to synthesize a wide array of capabilities and knowledge from its diverse donor models.
Key Characteristics
- Architecture: Based on the Gemma-4 family, specifically
google/gemma-4-26B-A4B-it. - Merge Method: Utilizes the
moe_dellatechnique, which is designed to blend experts effectively. - Context Length: Supports a substantial context window of 32768 tokens.
- Diverse Foundation: Incorporates models fine-tuned for various purposes, including instruction-following, creative writing, reasoning, and general knowledge.
Use Cases
This model is suitable for a broad spectrum of text generation tasks due to its composite nature. While it exhibits standard censorship and jailbreak resistance, advanced jailbreaking techniques or further decensoring efforts may be applied for specific needs. It is a versatile option for developers looking for a robust, merged model with a large context window.