|
|
--- |
|
|
license: apache-2.0 |
|
|
tags: |
|
|
- qwen3 |
|
|
- 256k context |
|
|
- thinking |
|
|
- instruct |
|
|
- reasoning |
|
|
- Brainstorm20x |
|
|
- Neo |
|
|
- Neo Imatrix |
|
|
base_model: |
|
|
- Qwen/Qwen3-4B-Thinking-2507 |
|
|
- Qwen/Qwen3-4B-Instruct-2507 |
|
|
--- |
|
|
|
|
|
<h2>Qwen3-4B-Instruct-2507-NEO-Imatrix-PlayGround-GGUF</h2> |
|
|
|
|
|
16 Neo Imatrix experimental configs (4B and 6B ggufs) for new Qwen 3 4B 2507 Instruct and "Thinking" 2507 too. |
|
|
|
|
|
All quants work. |
|
|
|
|
|
This repo is a test bed for each quant/config type, which will be followed by a single repo per winning "configs". |
|
|
|
|
|
Repo contains a "baseline" also of Instruct and Thinking as a reference - no "NEO" / 20-2 in the filename. |
|
|
|
|
|
6 Quants at 6B parameters/55 layers/607 tensors (3 Instruct, 3 Thinking) of Brainstorm 20x (denoted by 20-2 in the file name) too. |
|
|
|
|
|
All quants: 256k context. |
|
|
|
|
|
Model Sources [read more about model, settings and so...]: |
|
|
|
|
|
https://huggingface.co/Qwen/Qwen3-4B-Instruct-2507 |
|
|
|
|
|
https://huggingface.co/Qwen/Qwen3-4B-Thinking-2507 |
