kainatq/ksong-1-12b_v1_m3

TEXT GENERATIONConcurrent Unit Cost:1Model Size:12BQuant:FP8Context Size:32kTool Calling:SupportedPublished:Jun 24, 2026License:apache-2.0Architecture:Transformer Open Weights Featherless Exclusive Cold

The kainatq/ksong-1-12b_v1_m3 is a 12 billion parameter language model created by kainatq, built by merging Gryphe/Pantheon-RP-1.5-12b-Nemo and DarwinAnim8or/Trouper-12B using a slerp merge method. This model leverages the strengths of its base components, offering a balanced performance profile. It is designed for general language generation tasks, benefiting from the combined capabilities of its constituent models.

Loading preview...

Model Overview

The ksong-1-12b_v1_m3 is a 12 billion parameter language model developed by kainatq. It was created using the mergekit tool, specifically employing a spherical linear interpolation (slerp) method to combine two distinct base models.

Key Characteristics

  • Architecture: The model is a merge of Gryphe/Pantheon-RP-1.5-12b-Nemo and DarwinAnim8or/Trouper-12B.
  • Merge Method: Utilizes slerp (spherical linear interpolation) for combining the weights of the base models, with a t parameter of 0.5, indicating an equal blend.
  • Parameter Count: Features 12 billion parameters, making it a moderately sized model suitable for a range of applications.
  • Data Type: Configured to use bfloat16 for its computations, balancing performance and memory efficiency.

Potential Use Cases

This merged model is likely to inherit and combine the strengths of its constituent models. While specific optimizations are not detailed, models like Pantheon-RP often focus on role-playing and creative generation, and Trouper models typically aim for robust general language understanding and generation. Therefore, ksong-1-12b_v1_m3 could be well-suited for:

  • General text generation and completion.
  • Creative writing and storytelling.
  • Conversational AI and role-playing scenarios.
  • Tasks requiring a balanced blend of factual recall and imaginative output.