---
license: apache-2.0
tags:
- merge
- mergekit
- lazymergekit
- Roman0/Qwen3-4B-Instruct-2507-heretic
- Jackrong/GPT-5-Distill-Qwen3-4B-Instruct
- thinkwee/NOVER1-Qwen3-4B
---

# Qwen3-4B-Exp

Qwen3-4B-Exp is a merge of the following models using [mergekit](https://github.com/cg123/mergekit):
* [Roman0/Qwen3-4B-Instruct-2507-heretic](https://huggingface.co/Roman0/Qwen3-4B-Instruct-2507-heretic)
* [Jackrong/GPT-5-Distill-Qwen3-4B-Instruct](https://huggingface.co/Jackrong/GPT-5-Distill-Qwen3-4B-Instruct)
* [thinkwee/NOVER1-Qwen3-4B](https://huggingface.co/thinkwee/NOVER1-Qwen3-4B)

## 🧩 Configuration

```yaml
models:
  - model: Roman0/Qwen3-4B-Instruct-2507-heretic
    parameters:
      density: 0.3
      weight: 0.3
  - model: Jackrong/GPT-5-Distill-Qwen3-4B-Instruct
    parameters:
      density: 0.4
      weight: 0.4
  - model: thinkwee/NOVER1-Qwen3-4B
    parameters:
      density: 0.4
      weight: 0.4
merge_method: ties
base_model: Roman0/Qwen3-4B-Instruct-2507-heretic
parameters:
  normalize: true
dtype: float16


```