nlpguy/Qwen3.8-27B-Kimiko-2

VISIONPricing:Input $1.6 / Cached $0.15 / Output $12Concurrent Unit Cost:2Model Size:27BQuant:FP8Context Size:32kTool Calling:SupportedPublished:Aug 25, 2026Architecture:Transformer0.0K Featherless Exclusive Cold

nlpguy/Qwen3.8-27B-Kimiko-2 is a 27 billion parameter language model based on the Qwen3.8 architecture, created by nlpguy through a Karcher Mean merge of four distinct Qwen3.8-27B variants. This model integrates capabilities from Qwen/Qwen3.8-27B, TeichAI/Qwen3.8-27B-Fable-Distill, beyoru/Kiwen1.1-27B, and nlpguy/Qwen3.8-27B-Fimi-4, offering a consolidated performance profile. It is designed for general language tasks, leveraging the combined strengths of its merged components.

Loading preview...

Overview

nlpguy/Qwen3.8-27B-Kimiko-2 is a 27 billion parameter language model developed by nlpguy. It was created using the Karcher Mean merge method via mergekit, combining the strengths of several Qwen3.8-27B base models. This merging approach aims to synthesize the capabilities of its constituent models into a single, robust offering.

Models Merged

This model is a composite of four distinct Qwen3.8-27B variants, each potentially contributing unique characteristics:

  • Qwen/Qwen3.8-27B: The foundational Qwen model.
  • TeichAI/Qwen3.8-27B-Fable-Distill: Likely a distilled or fine-tuned version.
  • beyoru/Kiwen1.1-27B: Another specialized variant.
  • nlpguy/Qwen3.8-27B-Fimi-4: A previous iteration or related model from the same creator.

Technical Configuration

The merge process utilized a qwen chat template and bfloat16 data types for efficiency. The Karcher Mean method was applied across various layers of the language model and visual blocks, indicating a comprehensive integration of the source models' parameters. This detailed merging strategy suggests an effort to balance and enhance the overall performance across different model components.

Intended Use

As a merged model, nlpguy/Qwen3.8-27B-Kimiko-2 is suitable for a broad range of general-purpose language generation and understanding tasks, benefiting from the collective intelligence of its diverse base models. Its 27 billion parameters and 32768 token context length make it a capable choice for applications requiring substantial linguistic capacity.

Finetunes
1 models