metadata

license: apache-2.0
tags:
  - merge
  - mergekit
  - lazymergekit
  - janhq/Jan-v1-4B
  - huihui-ai/Huihui-Qwen3-4B-Thinking-2507-abliterated
  - minchyeom/Qwaifu
base_model:
  - huihui-ai/Huihui-Qwen3-4B-Thinking-2507-abliterated
  - janhq/Jan-v1-4B
  - minchyeom/Qwaifu

Qwen3-4B-Pro

Can Be Used For:

The model is designed for a range of text generation tasks and is particularly effective in the following areas:

Deep Thinking: Multi-step logical reasoning and problem-solving.
Roleplay: Ability to role-playing scenarios.
Creative Writing: Forms of creative text.
Coding: It has a strong capability in generating, completing, and debugging code.

Limitations:

As a 4B parameter model, it may not match the performance of much larger models on highly complex or specialized tasks.

🧩 Configuration

models:
  - model: janhq/Jan-v1-4B
    parameters:
      density: 0.4
      weight: 0.4
  - model: huihui-ai/Huihui-Qwen3-4B-Thinking-2507-abliterated
    parameters:
      density: 0.5
      weight: 0.5
  - model: minchyeom/Qwaifu
    parameters:
      density: 0.2
      weight: 0.2
merge_method: ties
base_model: huihui-ai/Huihui-Qwen3-4B-Thinking-2507-abliterated
parameters:
  normalize: true
dtype: float16