Modelpublic

Abiray/Qwopus3.6-27B-v1-preview-GGUF

sourceHugging Faceapache-2.0updated 5mo agoView on Hugging Face

7likes212downloads

Model Card

Qwopus3.6-27B-v1-preview - GGUF

This repository contains GGUF quantized formats of Jackrong/Qwopus3.6-27B-v1-preview.

Qwopus3.6-27B-v1-preview is an early preview reasoning model built on top of the Qwen3.6-27B multimodal base. It is heavily fine-tuned to deliver stronger reasoning quality, a stable answer structure, and more consistent long-form responses. It defaults to a "thinking" mode where reasoning is generated inside <think>...</think> tags prior to the final response.

Available Quantizations

The following quantization formats are provided to balance VRAM/RAM usage and model performance:

File Name	Quant Type	Description
`Qwopus3.6-27B-v1-preview-Q4_K_M.gguf`	Q4_K_M	Recommended. Excellent balance of quality and size.
`Qwopus3.6-27B-v1-preview-Q4_K_S.gguf`	Q4_K_S	Slightly smaller than Q4KM, minor quality trade-off.
`Qwopus3.6-27B-v1-preview-Q5_K_M.gguf`	Q5_K_M	Higher precision. Requires more RAM/VRAM.
`Qwopus3.6-27B-v1-preview-Q5_K_S.gguf`	Q5_K_S	Good balance for those who want Q5 precision with slightly lower memory footprint.
`Qwopus3.6-27B-v1-preview-Q6_K.gguf`	Q6_K	Near-unquantized quality, very large file size.
`Qwopus3.6-27B-v1-preview-Q8_0.gguf`	Q8_0	Highest quality, virtually indistinguishable from FP16.

Prompt Format

This model uses the standard Qwen chat template. By default, it operates in a reasoning mode. The output format generally follows:

text

<|im_start|>system
You are a helpful assistant.<|im_end|>
<|im_start|>user
[Your Prompt Here]<|im_end|>
<|im_start|>assistant
<think>
[Reasoning trace]
</think>
[Final Answer]<|im_end|>