mrcuddle/Nebula-Reasoner-12B

TEXT GENERATIONPricing:Input $0.87 / Cached $0.2 / Output $0.99Concurrent Unit Cost:1Model Size:12BQuant:FP8Context Size:32kTool Calling:SupportedPublished:Aug 17, 2026Architecture:Transformer0.0K Featherless Exclusive Cold

mrcuddle/Nebula-Reasoner-12B is a 12 billion parameter language model created by mrcuddle using the Arcee Fusion merge method. It combines ChaoticNeutrals/Mag-Mell-Reasoner-12B and Nitral-AI/CaptainErisNebula-12B-Chimera-v1.1, leveraging their strengths. This model is designed for reasoning tasks, offering a 32768 token context length for complex problem-solving and detailed analysis.

Loading preview...

Model Overview

mrcuddle/Nebula-Reasoner-12B is a 12 billion parameter language model developed by mrcuddle. It was created using the Arcee Fusion merge method, a technique designed to combine the strengths of multiple pre-trained models. This approach aims to produce a model with enhanced capabilities, particularly in reasoning.

Merge Details

This model is a fusion of two distinct base models:

  • ChaoticNeutrals/Mag-Mell-Reasoner-12B: This served as the primary base for the merge.
  • Nitral-AI/CaptainErisNebula-12B-Chimera-v1.1: This model was integrated with a weight of 0.5 during the fusion process.

The combination of these models suggests an optimization for tasks requiring logical inference and complex understanding. With a substantial context length of 32768 tokens, Nebula-Reasoner-12B is well-suited for processing and reasoning over extensive inputs.

Potential Use Cases

Given its architecture and the 'Reasoner' designation in its base models, Nebula-Reasoner-12B is likely to excel in:

  • Complex problem-solving
  • Detailed analytical tasks
  • Applications requiring deep contextual understanding