Naphula/Goetia-26B-A4B-v1.6

VISIONPricing:Input $0.4 / Cached $0.08 / Output $0.8Concurrent Unit Cost:2Model Size:26BQuant:FP8Context Size:32kTool Calling:SupportedPublished:Sep 10, 2026License:apache-2.0Architecture:Transformer0.0K Open Weights Featherless Exclusive Cold

Naphula/Goetia-26B-A4B-v1.6 is a 26 billion parameter language model based on the Gemma4 architecture, created by Naphula using the moe_della merge method. This version incorporates updated Orion and Pantheon models and an upgraded lm_head, replacing Dark Goetia with Boulesis v1. It is designed for general language tasks, leveraging a diverse blend of over 25 specialized models to enhance its capabilities.

Loading preview...

Goetia 26B A4B v1.6 Overview

Naphula/Goetia-26B-A4B-v1.6 is a 26 billion parameter language model built upon the Gemma4 architecture, utilizing the moe_della merge method. This iteration, a successor to v1.5, integrates newer versions of Orion and Pantheon models and features an upgraded lm_head. Notably, it replaces Dark Goetia with Boulesis v1, aiming for improved performance and characteristics.

Key Characteristics

  • Architecture: Based on Gemma4ForConditionalGeneration with google/gemma-4-26B-A4B-it as the base model.
  • Merge Method: Employs the moe_della merge method, a technique designed for combining multiple pre-trained models effectively.
  • Diverse Model Integration: Merges over 25 distinct models, including specialized versions like ApocalypseParty/G4-26B-SFT-6, AuriAetherwiing/G4-26B-A4B-Musica-v1, Gryphe/Pantheon-Reasoning-26B-A4B-1.1-V2, and SubMaroon/Boulesis-26B-A4B, among many others.
  • Context Length: Supports a context length of 32768 tokens.

Considerations

  • Censorship: The model is not uncensored and exhibits standard refusals and jailbreak resistance. Advanced jailbreaks or decensoring tools like Heretic may be used to bypass these limitations.

Finetunes
1 models