Nelathan/Qwen3-8B-rooted

Qwen3-rooted

This is a merge of pre-trained language models created using mergekit.

Merge Details

Merge Method

This model was merged using the Multi-SLERP merge method using Qwen/Qwen3-8B-Base as a base.

Models Merged

The following models were included in the merge:

• deepseek-ai/DeepSeek-R1-0528-Qwen3-8B
• allura-org/remnant-qwen3-8b
• ArliAI/DS-R1-Qwen3-8B-ArliAI-RpR-v4-Small
• Qwen/Qwen3-8B
• allura-org/Q3-8B-Kintsugi

Configuration

The following YAML configuration was used to produce this model:

merge_method: multislerp
parameters:
  normalize: true
  int8_mask: true
dtype: bfloat16

base_model: Qwen/Qwen3-8B-Base
tokenizer_source: deepseek-ai/DeepSeek-R1-0528-Qwen3-8B
models:
  - model: deepseek-ai/DeepSeek-R1-0528-Qwen3-8B
    parameters:
      weight: 0.5
  - model: ArliAI/DS-R1-Qwen3-8B-ArliAI-RpR-v4-Small
    parameters:
      weight: 0.5
  - model: Qwen/Qwen3-8B
    parameters:
      weight: 0.5
  - model: allura-org/Q3-8B-Kintsugi
    parameters:
      weight: 0.5
  - model: allura-org/remnant-qwen3-8b
    parameters:
      weight: 0.33