license: apache-2.0
tags:
- gguf
- qwen
- qwen4
- quantized
- mixed-precision
- uncensored
- abliterated
- consumer-gpu
language: - en
- zh
- code
pipeline_tag: text-generation
Qwen4-27B · XQ, abliterated — reserved, no weights yet
This repository contains no model files and is not ready for inference. It reserves the name for
Qwen4-27B with the refusal direction edited out of the weights, published in Indra's Mirror's
mixed-precision XQ format, once Alibaba releases Qwen4-27B.
Follow the repository to be notified when the files land.
What will be published
| base | Qwen/Qwen4-27B, the official BF16 release |
| edit | refusal-direction removal applied to the full-precision weights before quantisation |
| format | XQ, mixed-precision GGUF, several sizes from a near-lossless tier down to small-GPU tiers |
| reference | an unedited abliterated high-precision GGUF, so every XQ file can be measured against it |
| extras | whatever projector or auxiliary files the official release ships, copied unchanged |
Model card, per-file sizes and the measurements that justify each size will replace this page when
the files are uploaded.
Why a reserved repository
Qwen4-27B has not been released. Several "Qwen4" repositories already exist on the Hub with no
weights behind them; this one says so plainly rather than implying otherwise.
When the base model is out, the plan is the usual shape: a size ladder rather than one file, each
tier documented with its own size and divergence from the unedited high-precision reference, so you
can pick by hardware instead of by guesswork.
Status
- Qwen4-27B released by Alibaba
- weights downloaded and verified against the published hashes
- refusal-direction edit fitted and checked on held-out prompts
- XQ size ladder produced
- measurements and model card published
Licence
Follows the Qwen4-27B release licence. Nothing here overrides it.