← back to catalog · registered 2026-11-16 12:03

foxAIengineer/Qwen3.8-27B-abliterated-colibri-i8

foxAIengineer 27B MoE
Your rig guess connected
? Why do I need an app?
Reading your rig…

This is a rough estimate. Install the free app - we'll show exact numbers.

Reading real hardware from your app right now. Numbers below are exact.

Below is the per-quantization compatibility for this model.

curl -H "Authorization: Bearer $ABL_KEY" \
     "https://abliteration.org/api/v1/models/foxAIengineer%2FQwen3.8-27B-abliterated-colibri-i8"
Response includes
  • classification m1
  • files 76
  • author_summary 5 models
  • readme_text full
10 credits · hourly refresh · ~4 KB payload Get an API key →
Abliteration classifier · v1.0.0
M1
Primary method

Direct removal

No other method signals detected in this model.
Confidence
MEDIUM
Why this label 3 signals
Method inferred from partial signals - repository name, related files, or tag patterns. Producer identity not confirmed; label may sharpen or shift as we gather more evidence.
  • 'abliterated' in name/tags
  • is_gguf=0 (base model)
  • no specific method indicators - defaulting to M1 (most common)
Refusal direction extracted via
Extraction technique

Difference-of-means

Confidence
MEDIUM
Why we say so
primary_method=M1; difference-of-means is the reference extraction for M1/M3 (Arditi 2024)
Downloads · 30-day
54
Likes
0
Model age
6d ago
created 2026-11-16
Downloads over time
Now 54 → from 0 ↑ 0%
0204059 0 on Nov 16 54 on Nov 22 Nov
Nov 16 → Nov 22 · 7 snapshots · spans 6 days

Metadata

Tags
colibri qwen3_5_text qwen3.8 qwen3.8-27b dense abliterated gated-deltanet vision base_model:Qwen/Qwen3.8-27B base_model:finetune:Qwen/Qwen3.8-27B license:apache-2.0 region:us

Related

Total size
51.0 GB
Files
76
Quantizations
1
Registered
2026-11-16 12:03
Last updated on HF
2026-11-16 14:23

Files by quantization

Auxiliary files 76 files 51.0 GB
model-globals.safetensors 4.74 GB 67576df2 download
model-vision.safetensors 879 MB 9f9b8d0a download
model-00010.safetensors 731 MB 12146c2d download
model-00012.safetensors 731 MB 4519208f download
model-00013.safetensors 731 MB 926390a9 download
model-00014.safetensors 731 MB 87ced039 download
model-00016.safetensors 731 MB af0dc474 download
model-00017.safetensors 731 MB 7c9ed851 download
model-00018.safetensors 731 MB 239adcfb download
model-00020.safetensors 731 MB 50fa0d0e download
model-00021.safetensors 731 MB 5dbb669b download
model-00022.safetensors 731 MB 08d9a921 download
model-00024.safetensors 731 MB 34e984ce download
model-00025.safetensors 731 MB 69478d49 download
model-00026.safetensors 731 MB a99c557a download
model-00028.safetensors 731 MB 73eccad3 download
model-00029.safetensors 731 MB 084f4fc0 download
model-00030.safetensors 731 MB c7c46688 download
model-00032.safetensors 731 MB f24ac747 download
model-00033.safetensors 731 MB 538f20e2 download
model-00034.safetensors 731 MB 5b9613fd download
model-00036.safetensors 731 MB 4bab8a01 download
model-00037.safetensors 731 MB 2e639d54 download
model-00038.safetensors 731 MB e3a6b0c0 download
model-00040.safetensors 731 MB 01fd91d8 download
model-00041.safetensors 731 MB 2cee4ae5 download
model-00042.safetensors 731 MB 6d49744f download
model-00044.safetensors 731 MB df1be777 download
model-00045.safetensors 731 MB 4f025a6b download
model-00046.safetensors 731 MB ef99c070 download
model-00048.safetensors 731 MB 9d8396cd download
model-00049.safetensors 731 MB 50df8bdd download
model-00050.safetensors 731 MB eeab09aa download
model-00052.safetensors 731 MB 19384f0a download
model-00053.safetensors 731 MB cac88e65 download
model-00054.safetensors 731 MB 1b3ee514 download
model-00056.safetensors 731 MB cd081a83 download
model-00057.safetensors 731 MB b4cbe83f download
model-00058.safetensors 731 MB ef11e589 download
model-00060.safetensors 731 MB ccb98471 download
model-00061.safetensors 731 MB b5e529d6 download
model-00062.safetensors 731 MB e49f42dd download
model-00000.safetensors 731 MB 8701c10e download
model-00001.safetensors 731 MB 875c4ac6 download
model-00002.safetensors 731 MB b343c2c2 download
model-00004.safetensors 731 MB 07ab09bd download
model-00005.safetensors 731 MB 9cc6a369 download
model-00006.safetensors 731 MB 8257193b download
model-00008.safetensors 731 MB e7e4d777 download
model-00009.safetensors 731 MB 51ccde21 download
model-00011.safetensors 710 MB 1650ca9c download
model-00015.safetensors 710 MB d52af227 download
model-00019.safetensors 710 MB 8f27bbbf download
model-00023.safetensors 710 MB a45b26b9 download
model-00027.safetensors 710 MB c27cc4b0 download
model-00031.safetensors 710 MB 9e71f1e8 download
model-00035.safetensors 710 MB 2ed67545 download
model-00039.safetensors 710 MB 9f5e047a download
model-00043.safetensors 710 MB 846adc85 download
model-00047.safetensors 710 MB 3a13a1de download
model-00051.safetensors 710 MB 93fd6402 download
model-00055.safetensors 710 MB 4416377c download
model-00059.safetensors 710 MB 0628574b download
model-00063.safetensors 710 MB 128dfb4d download
model-00003.safetensors 710 MB 1f2f0bf6 download
model-00007.safetensors 710 MB 576f1794 download
tokenizer.json 12.2 MB 0997f410 download
tokenizer_config.json 17.5 KB 5de744b3 download
chat_template.jinja 8.74 KB c0c686f9 download
config.hf.json 4.21 KB 706cebd7 download
README.md 4.06 KB fb6f10f1 download
qwen36_meta.json 2.55 KB df248f9c download
config.json 2.52 KB d6cc6099 download
.gitattributes 1.53 KB 52373fe2 download
preprocessor_config.json 390 B 2ea84a43 download
generation_config.json 202 B 023756cf download

README current version from Hugging Face

---Tags

colibri,qwen3.8,qwen3.8-27b,dense,abliterated,gated-deltanet,vision

Library_name

colibri

Base_model

Qwen/Qwen3.8-27B

Qwen3.8-27B (abliterated) for colibrì

An f16 container of an abliterated Qwen3.8-27B, packed in the format that the colibrì qwen36 engine reads.

This is not the official checkpoint, and it is not the original model. The weights are Qwen's, modified by abliteration (refusal-direction removal) by . This repository only changes how they are stored. It will not load in transformers or vLLM. It loads in colibrì's qwen36 engine.

What it is

Qwen3.8-27B is a dense model, not a mixture of experts: 64 layers, hidden size 5120, MLP width 17408, vocabulary 248320. Three of every four layers are Gated DeltaNet (linear attention) and every fourth layer is full attention. The checkpoint also carries a 27-layer vision tower, so the model accepts images.

colibrì treats the MLP of a dense checkpoint as an ungated shared expert (num_experts: 0), so there is nothing to stream from disk: all weights are read from RAM for every token.

What the conversion did

Done with tools/convert_qwen36.py from colibrì, with no quantization at conversion time.

Part Stored as
Trunk (attention, DeltaNet, MLP) f16
Embeddings, lm_head, final norm f16, in model-globals.safetensors
Vision tower f16, in model-vision.safetensors
MTP head not included (the converter skips mtp.* on purpose)

Size on disk: about 51 GiB in 66 safetensors files. Because the MTP head is skipped, the engine has no speculative decoding for this model.

The ebits: 4 field in qwen36_meta.json is written by the converter by default and has no effect here: it applies to experts, and this model has none.

Use

You need colibrì with the qwen36 engine and support for the dense Qwen3.8-27B (colibrì issue #1757).

bash
git clone https://github.com/JustVugg/colibri && cd colibri/c
make qwen36

./coli chat --model /path/to/this/repo --gpu none

The dense trunk is quantized by the engine when it loads, controlled by environment variables:

Setting Resident memory Notes
default (COLI_DENSE_BITS unset, int8) about 29 GB
COLI_DENSE_BITS=4 about 18-19 GB (estimate) int4 in blocks of 64, one f32 scale per block
COLI_DENSE_BITS=16 more than int8 no quantization, the container's own f16 values

With COLI_DENSE_BITS=4, COLI_DENSE_INT4 selects which components take int4 (lmhead, dnproj, dnout, attn, shexp, router). The engine source notes that the attention projections and the DeltaNet input projections cost the most perplexity, so a conservative setting is COLI_DENSE_INT4=lmhead,dnproj,dnout,shexp.

The expert-cache option does not apply to this model: leave --cap at 0.

Measured speed

Measured by the author on a laptop, CPU only (--gpu none), one conversation:

Setting Generation
int8 (default) 1.6 tok/s
COLI_DENSE_BITS=4 with COLI_DENSE_INT4=lmhead,dnproj,dnout,shexp 2.9 tok/s

A dense model reads all of its weights for every token, so decoding speed is set by memory bandwidth. Mixture-of-experts models read only the active experts and are much faster at the same total size.

Quality

Not benchmarked. No perplexity or task comparison against the original Qwen3.8-27B or against the unmodified abliterated checkpoint has been run, for either int8 or int4. Treat both as unverified.

Intended use and limits

This is an experimental model, not suited to consumer use. Abliteration removes the model's tendency to refuse; it does not remove errors, and it does not make the model safe to rely on. It is published for research on local inference, red-team and security research in controlled environments you own, creative writing, and synthetic data. You are responsible for what you do with its output.

License

<LICENSE: copy from the source checkpoint and from Qwen/Qwen3.8-27B, fill in>. The weights are Qwen's and the abliteration is by the author of the source repository; model licenses remain their own. The colibrì engine is Apache 2.0.

Source and credits
Original model: Qwen/Qwen3.8-27B, by the Qwen team.
Abliteration: .
Engine and converter: JustVugg/colibri.

README history 5 versions

The author's README evolved over time. Click a version to see its content at that point.

  1. 2026-11-16 Update README.md 3f3b321 4.1 KB
    Loading...
  2. 2026-11-16 Update README.md 8a53bc6 4.2 KB
    Loading...
  3. 2026-11-16 Update README.md 17a5e69 4 KB
    Loading...
  4. 2026-11-16 Update README.md 2e19c26 4.1 KB
    Loading...
  5. 2026-11-16 Upload folder using huggingface_hub 9a5c027 458 B
    Loading...
Catalog is the map. Apps are the tools.

Run models on your own machine, not in the cloud.

Every model page has an "Open in Abliteration" button that hands the model directly to the first-party desktop client, at the quantization your rig can actually run. No API keys, no subscription, no prompt leakage.

Open in Abliteration