grimjim/Equatorium-v3-12B

TEXT GENERATIONPricing:Input $0.87 / Cached $0.2 / Output $0.99Concurrent Unit Cost:1Model Size:12BQuant:FP8Context Size:32kTool Calling:SupportedPublished:Feb 22, 2026License:apache-2.0Architecture:Transformer0.0K Open Weights Featherless Exclusive Cold

grimjim/Equatorium-v3-12B is a 12 billion parameter language model created by grimjim using the MergeKit framework. This model is a merge of several pre-trained models, including Delta-Vector/Rei-V3-KTO-12B and grimjim/AbMagnolia-v1-12B, with grimjim/mistralai-Mistral-Nemo-Base-2407 as its base. It was developed to provide stability and coherence, even under high temperature settings, making it suitable for general text generation tasks.

Loading preview...

grimjim/Equatorium-v3-12B Overview

This model is a 12 billion parameter language model developed by grimjim, leveraging the MergeKit framework. It addresses stability issues found in prior merges by adopting a "Magnum-adjacent" approach, aiming for consistent and coherent output.

Key Characteristics

  • Architecture: A merged model built upon grimjim/mistralai-Mistral-Nemo-Base-2407 as its foundation.
  • Merge Method: Utilizes the Task Arithmetic merge method, combining components from Delta-Vector/Rei-V3-KTO-12B, grimjim/AbMagnolia-v1-12B, and grimjim/mistralai-Mistral-Nemo-Instruct-2407.
  • Stability: Designed for improved stability and coherence, tested to perform well under temp=1.0 and minP=0.02.

Use Cases

  • General Text Generation: Suitable for a wide range of text generation tasks where stability and coherence are important.
  • Exploration: Can be used by developers experimenting with merged models and their performance characteristics under varying generation parameters.

Finetunes
1 models