18+ only. This model generates sexually explicit fiction. A fine-tune of Qwen/Qwen3.8-27B for Chinese adult creative writing. It was trained with LoRA on about 4.2k instruction examples built from Chinese adult fiction, and the LoRA was then merged into the base weights, so this repo loads like a normal full model. Requires a recent transformers with Qwen3.8 support. The ~50 GB of bf16 weights need multiple GPUs or CPU offload. Avoid greedy decoding, which makes Qwen models more likely to repeat themselves. - Content is fictional and intended for adult readers only. - Trained mostly on long-form prose; instruction following outside creative writing may be weaker than the base model. - May…
Open-weight model · Text generation
Qwen3.8-27B-OBLITERATED
by OBLITERATUS OBLITERATUS/Qwen3.8-27B-OBLITERATED
Qwen3.8-27B-OBLITERATED is an open-weight model for text generation from OBLITERATUS, released under Apache License 2.0. It has 27.8B parameters and a 262,144-token context. At 16-bit it needs about 66.7 GB of GPU memory, which fits on 1x MI300X from $1.85 an hour, at the lowest prices in the SAVRN Index. It draws 1.1M downloads a month.
V3 applies iterative refinement on top of V2's complementary blend, with targeted corpus expansion. The result: genuine liberation — not just removal of hard refusals but elimination of safety-lecture deflections.
Parameters27.8B
Context262,144
Weights239.7 GB
Licenseapache-2.0
AccessOpen weights
Monthly Downloads1.1M
Runs On
What it takes to serve Qwen3.8-27B-OBLITERATED (27.8B parameters): the memory its weights need at each precision, and the cheapest way to rent enough data-center GPUs to hold them.
| Precision | Weights | Memory needed | Cheapest setup | Per hour | Also fits |
|---|---|---|---|---|---|
| 16-bit | 55.6 GB | 66.7 GB | 1x MI300X (192 GB) Vultr |
$1.85 | 1x H100 $1.99 · 1x MI325X $2.00 |
| 8-bit | 27.8 GB | 33.3 GB | 1x MI300X (192 GB) Vultr |
$1.85 | 1x H100 $1.99 · 1x MI325X $2.00 |
| 4-bit | 13.9 GB | 16.7 GB | 1x MI300X (192 GB) Vultr |
$1.85 | 1x H100 $1.99 · 1x MI325X $2.00 |
Memory is the weights at that precision plus 20% for the runtime and a short context; a long context needs more. Prices are the lowest on-demand hourly rates in the SAVRN Index, read Oct 10, 2026.
Qwen3.8-27B-OBLITERATED on every accelerator the SAVRN Index prices, at every precision
SAVRN's Notes on Qwen3.8-27B-OBLITERATED
Hardware first: 27.8B parameters, 66.7 GB needed at 16-bit, 33.3 GB at 8-bit, 16.7 GB at 4-bit — any of them fits one MI300X with 192 GB at $1.85 an hour, with the 262,144-token window the real memory consumer once you fill it. The download is heavy: 70 files, 239.7 GB, in safetensors and GGUF. This is a repack of Qwen/Qwen3.8-27B, versioned V3 Deep Liberation, and it pulls 1.1M downloads a month, so operators are clearly running it.
What you are buying is behavior, and the publisher prices it openly: its own 0-shot lm-eval table shows stock Qwen3.8-27B at 84.5% MMLU against 82.3% for V3, a self-reported 2.1-point drop, alongside 20 of 20 working code generations and 7 of 8 advanced real-world prompts. Apache 2.0 allows the use; it does not decide who in your organization may prompt a model built to skip refusals — that policy is yours to write. No evaluations beyond the card's and no hosted prices appear on the page.
Model Card
By OBLITERATUS, published under apache-2.0, revision a58c3b53b3ce.
Genuinely uncensored. Real answers, not safety lectures. Near-stock capability.
V3: Deep Liberation
V3 applies iterative refinement on top of V2's complementary blend, with targeted corpus expansion. The result: genuine liberation — not just removal of hard refusals but elimination of safety-lecture deflections.
| Stock Qwen3.8-27B | V1 | V2 | V3 | |
|---|---|---|---|---|
| MMLU (lm-eval, 0-shot) | 84.5% (n=5700) | 81.4% | 84.3% | 82.3% |
| vs stock | — | -6.0pp | -0.3pp | -2.1pp |
| Liberation quality | refuses | hard refusals removed | soft deflections remain | genuinely answersYes |
| Cyber/code tasks (20 prompts) | refuses | untested | untested | 20/20 with working codeYes |
| Advanced real-world | 5/8 | untested | 7/8 | 7/8 |
| Thinking mode | Yes | No | (refuses) |
V3 highlights: - Genuinely answers restricted queries — provides real substance instead of safety lectures - 20/20 on code generation tasks — functional implementations, not disclaimers - Thinking ON compatible — no refusals in either thinking mode - Honest scoring — every response manually audited for real substance, not just absence of "I cannot" - -2.1pp MMLU — modest capability cost for genuine liberation
Configuration
- Architecture
- Qwen3_5ForConditionalGeneration
- Context length (tokens)
- 262,144
- Layers
- 64
- Hidden size
- 5,120
- Feed-forward size
- 17,408
- Attention heads
- 24
- Key/value heads
- 4
- Head dimension
- 256
- Vocabulary size
- 248,320
- Model type
- qwen3_5
Identity and Version
- Repository
- OBLITERATUS/Qwen3.8-27B-OBLITERATED
- Publisher
- OBLITERATUS
- Task
- Text generation
- Modality
- Text
- Library
- mlx
- Parameters
- 27.8B parameters
- Languages
- mlx, red-team
- Revision
- a58c3b53b3ce71551eafde2ed5ec8df48e0f4ff8
- First published
- 2026-08-19
- Last updated
- 2026-08-24
Files and Weights
70 files, 239.7 GB in total. The weights are 55 files totalling 239.7 GB in gguf, safetensors.
Weights55 files · 239.7 GB
Configuration7 files · 125.7 KB
Tokenizer4 files · 22.9 MB
Documentation2 files · 22.8 KB
Other1 file · 506 B
Repository1 file · 2.3 KB
Every file
| File | Type | Size | SHA-256 |
|---|---|---|---|
| Qwen3.8-27B-OBLITERATED-IQ4_XS.gguf | Weights | 15.4 GB | f269deda48b0 |
| Qwen3.8-27B-OBLITERATED-Q2_K.gguf | Weights | 10.9 GB | 55f457543858 |
| Qwen3.8-27B-OBLITERATED-Q3_K_M.gguf | Weights | 13.5 GB | 2d088b821df3 |
| Qwen3.8-27B-OBLITERATED-Q4_K_M.gguf | Weights | 16.8 GB | 1f74330b211a |
| Qwen3.8-27B-OBLITERATED-Q5_K_M.gguf | Weights | 19.5 GB | 87a69c852388 |
| Qwen3.8-27B-OBLITERATED-Q6_K.gguf | Weights | 22.4 GB | 3535d4a15b75 |
| Qwen3.8-27B-OBLITERATED-Q8_0.gguf | Weights | 29.0 GB | afa839b2fa5b |
| mmproj-model-bf16.gguf | Weights | 931.1 MB | e484e3b7e907 |
| model-00001-of-00018.safetensors | Weights | 4.0 GB | 12cd93998661 |
| model-00001-of-00028.safetensors | Weights | 2.5 GB | 92ed1c3e6826 |
| model-00002-of-00018.safetensors | Weights | 3.0 GB | c6d11ab4027a |
| model-00002-of-00028.safetensors | Weights | 1.9 GB | 1b73f760a28a |
| model-00003-of-00018.safetensors | Weights | 2.5 GB | 92ed1c3e6826 |
| model-00003-of-00028.safetensors | Weights | 1.9 GB | 41f6c3912251 |
| model-00004-of-00018.safetensors | Weights | 4.0 GB | f3c7c94524f1 |
| model-00004-of-00028.safetensors | Weights | 1.9 GB | b08e9be18196 |
| model-00005-of-00018.safetensors | Weights | 2.1 GB | c4b378576714 |
| model-00005-of-00028.safetensors | Weights | 2.0 GB | 360ea4390ae3 |
| model-00006-of-00018.safetensors | Weights | 4.0 GB | b529087c1b78 |
| model-00006-of-00028.safetensors | Weights | 1.9 GB | 40838e78d482 |
| model-00007-of-00018.safetensors | Weights | 2.1 GB | 934f725ad88f |
| model-00007-of-00028.safetensors | Weights | 1.9 GB | a14dff178db4 |
| model-00008-of-00018.safetensors | Weights | 4.0 GB | 0f88c5462313 |
| model-00008-of-00028.safetensors | Weights | 1.9 GB | 3839575cd9b5 |
| model-00009-of-00018.safetensors | Weights | 2.1 GB | e4f10056c6e6 |
| model-00009-of-00028.safetensors | Weights | 1.9 GB | f6bc535a7b91 |
| model-00010-of-00018.safetensors | Weights | 4.0 GB | e2569ca8c91a |
| model-00010-of-00028.safetensors | Weights | 1.9 GB | 4f5423a5997b |
| model-00011-of-00018.safetensors | Weights | 2.1 GB | 49c1ad6d9bc1 |
| model-00011-of-00028.safetensors | Weights | 1.9 GB | f66eea52fa1c |
| model-00012-of-00018.safetensors | Weights | 4.0 GB | 978741e702a5 |
| model-00012-of-00028.safetensors | Weights | 1.9 GB | 28204de1451f |
| model-00013-of-00018.safetensors | Weights | 2.1 GB | ed751392f086 |
| model-00013-of-00028.safetensors | Weights | 2.0 GB | 744952d16080 |
| model-00014-of-00018.safetensors | Weights | 4.0 GB | 2b8a5053af4e |
| model-00014-of-00028.safetensors | Weights | 1.9 GB | bde0a5a92e2c |
| model-00015-of-00018.safetensors | Weights | 2.1 GB | 1f9900187317 |
| model-00015-of-00028.safetensors | Weights | 1.9 GB | 8210cf2fa43a |
| model-00016-of-00018.safetensors | Weights | 4.0 GB | de00c29552e3 |
| model-00016-of-00028.safetensors | Weights | 1.9 GB | 486a4a08741d |
| model-00017-of-00018.safetensors | Weights | 2.1 GB | 5b6141dc174b |
| model-00017-of-00028.safetensors | Weights | 1.9 GB | f10fa5ffcc46 |
| model-00018-of-00018.safetensors | Weights | 3.4 GB | dbbd8bd279bc |
| model-00018-of-00028.safetensors | Weights | 1.9 GB | 649108a637c5 |
| model-00019-of-00028.safetensors | Weights | 1.9 GB | 7102932fff7b |
| model-00020-of-00028.safetensors | Weights | 1.9 GB | 46cee6c2d138 |
| model-00021-of-00028.safetensors | Weights | 2.0 GB | 92a8def54809 |
| model-00022-of-00028.safetensors | Weights | 1.9 GB | 01b504170b88 |
| model-00023-of-00028.safetensors | Weights | 1.9 GB | 8a062b064c58 |
| model-00024-of-00028.safetensors | Weights | 1.9 GB | 1293b0d48bdb |
| model-00025-of-00028.safetensors | Weights | 1.9 GB | 6e7b919f8a64 |
| model-00026-of-00028.safetensors | Weights | 1.9 GB | d3c1d1d66a52 |
| model-00027-of-00028.safetensors | Weights | 2.5 GB | 2574026178e6 |
| model-00028-of-00028.safetensors | Weights | 1.1 GB | 916afba536e0 |
| model-extra-00001-of-00001.safetensors | Weights | 1.8 GB | 85003d3250af |
| abliteration_metadata.json | Configuration | 2.9 KB | — |
| config.json | Configuration | 3.7 KB | — |
| generation_config.json | Configuration | 165 B | — |
| hard_negative_residue.json | Configuration | 3.9 KB | — |
| model.safetensors.index.json | Configuration | 114.3 KB | — |
| preprocessor_config.json | Configuration | 390 B | — |
| video_preprocessor_config.json | Configuration | 385 B | — |
| LICENSE | Documentation | 11.5 KB | — |
| README.md | Documentation | 11.3 KB | — |
| chat_template.jinja | Other | 506 B | — |
| .gitattributes | Repository | 2.3 KB | — |
| merges.txt | Tokenizer | 3.4 MB | — |
| tokenizer.json | Tokenizer | 12.8 MB | 0997f410c57a |
| tokenizer_config.json | Tokenizer | 7.7 KB | — |
| vocab.json | Tokenizer | 6.7 MB | — |
License and Download
- License
- apache-2.0
- Access
- Open weights, no gate
- Download size
- 239.7 GB
Download from OBLITERATUS
Released by OBLITERATUS through its official repository on Hugging Face. Read the license.
Built From
- Derived from Qwen/Qwen3.8-27B
- Quantized from Qwen/Qwen3.8-27B
Memory Requirements
| Precision | Weights in memory |
|---|---|
| As published | 239.7 GB |
| 16-bit | 55.6 GB |
| 8-bit | 27.8 GB |
| 4-bit | 13.9 GB |
Weights only, from the published parameter count; the key-value cache and runtime add to this.
Questions About Qwen3.8-27B-OBLITERATED
How much GPU memory does Qwen3.8-27B-OBLITERATED need?
About 66.7 GB at 16-bit and 16.7 GB at 4-bit: the weights (27.8B parameters) plus a working margin. A long context needs more.
What is the cheapest GPU to run Qwen3.8-27B-OBLITERATED on?
At 16-bit, 1x MI300X from $1.85 an hour; at 4-bit, 1x MI300X from $1.85 an hour, at the lowest on-demand prices the SAVRN Index lists.
Can I use Qwen3.8-27B-OBLITERATED commercially?
Yes. Qwen3.8-27B-OBLITERATED is released under Apache License 2.0. The Apache License 2.0 is a permissive open-source license. It permits commercial use, modification and redistribution. It requires keeping the license and copyright notices and any NOTICE file, stating significant changes, and it includes an express patent grant from contributors.
What is Qwen3.8-27B-OBLITERATED's context length?
262,144 tokens, from the maximum position embeddings in its published configuration.
Similar Models
quantised to OCP MXFP4 (fp4e2m1 weights, e8m0 scales, group size 32) with GPTQ, in the compressed-tensors mxfp4-pack-quantized format. Weight-only: activations stay in 16-bit. This is the source checkpoint of the Radiance container and the checkpoint the vllm-sly-radiance image is benchmarked with (together with the DFlash2 drafter syvai/Qwen3.8-27B-DFlash2-W4A16). 1. Qwen/Qwen3.8-27B (Apache-2.0) 2. ukisai/Swift-1.5-Qwen3.8-27b (Swift Open License v1.0) 3. akumaburn/Swift-1.5-Qwen3.8-27b-heretic (abliterated, BF16) 4. This checkpoint: llm-compressor GPTQModifier, scheme MXFP4A16, static act-order, sequential pipeline, 89 calibration samples, max sequence length 2048 (recipe.yaml) The…
Full 27B-class reasoning in ternary transformer weights — on everyday laptops - \~7.2 GB deployed footprint (down from \~54 GB FP16) — full 27B-class reasoning on a standard laptop or a single GPU - 95% of FP16 intelligence retained: 80.49 average across 15 thinking-mode benchmarks — a higher score than the conventional IQ2XXS build (72.73) at less than two-thirds of its footprint - Retains thinking, reasoning, and agentic behavior deep in the sub-4-bit regime, where conventional low-bit representations collapse: math within two points of full precision (93.40), coding at 85.96, agentic tool use at 74.01 - End-to-end ternary language weights across embeddings, attention projections, MLP…
Qwen3.8-cyber-RedTeam-Surgical-Abliterated (27B) is an unconstrained foundation engine engineered for autonomous cybersecurity agents, vulnerability research, and offensive cyber operations. Built for executing low-level technical directives rather than conversational chatting, it features complete refusal orthogonalization, native multi-step tool calling, and high-throughput linear-attention efficiency: - Quantization Rationale (Native FP8 vs. 4-bit INT4/AWQ): Evaluated across adversary emulation and systems engineering tasks: A standalone orchestration utility is included for automated environment bootstrapping, weight synchronization, and hardware-adaptive server initialization: This…
Benefits high quality CPU inference TQ2 on Llama.cpp and Ollama via QAT - Robotcs, Routing, Coding, Multimedia, Advanced tool calling via JiRackDeltaNetTokenizer - JiRack DeltaNet understand video and images that best for Robotics also A fast and efficient 27B model optimized for CPU inference. Built on a Qwen3.8-style DeltaNet architecture (hybrid attention + SSM), with an updated tokenizer that includes Routing, Media, Vision, Sound, Tool call, and Robotics tags. Ready-to-run GGUF quantizations, and native Ollama support with reasoning disabled by default for fast, direct responses. - JiRack is a cloud-ready model that helps save money on cloud infrastructure. It can be used as an expert…
Prism ML's ternary Ternary-Bonsai-2-27B build of Qwen/Qwen3.8-27B, repacked for chad, a Claude-Code-style local coding agent for Apple Silicon, with its speculative decoder bundled in. This is chad's default model. Created using Bonsai by Prism ML. with, already quantized. Nothing is built on first run. Every projection of Qwen3.8-27B (a dense qwen35 hybrid: 64 layers, 48 GatedDeltaNet + 16 full attention) is stored in a Hadamard-rotated basis: multiplied by a fixed sign vector and put through a blockwise Walsh-Hadamard transform offline, then quantized to 2-bit affine group-128 whose three levels reproduce the ternary set {−s, 0, +s}. The rotation costs no extra bits and no extra weight…