license: apache-2.0
base_model: noon-at-cgn/Qwen3.8-27B-Uncensored-W4A16-AutoRound
base_model_relation: quantized
library_name: transformers
pipeline_tag: image-text-to-text
language:

• en
• zh
  tags:
• qwen3.8
• qwen3_5
• uncensored
• abliterated
• compressed-tensors
• autoround
• gptq
• w4a16
• int4
• mtp
• speculative-decoding
• vllm
• rtx-3090
• embed-int4
• int4-embeddings

Qwen3.8-27B-Uncensored-W4A16-RTX3090-MTP4

A standalone, serving-optimized derivative of
noon-at-cgn/Qwen3.8-27B-Uncensored-W4A16-AutoRound,
built and measured for low-latency MTP speculative decoding on one 24 GB RTX
3090-class GPU using
syv-ai/qwen38-27b-rtx3090.

This repository contains the complete checkpoint. It does not require the
upstream weight repository after download.

[!NOTE]
Variant — INT4 token embeddings (this repository). This is a re-quantized
derivative of
protogeni/Qwen3.8-27B-Uncensored-W4A16-RTX3090-MTP4
in which only the token-embedding group (model.language_model.embed_tokens)
is reduced from INT8 to INT4 (same group-128 symmetric pack-quantized layout).
Every other tensor, shard, and file is byte-identical to that source. See the
“Variant: INT4 token embeddings” section below for the exact change, packing
layout, and measured precision. The protogeni repository remains the untouched
original (INT8 embedding).

[!WARNING]
This checkpoint inherits substantial safety-alignment removal from its
abliterated upstream model. It can comply with harmful, unethical,
offensive, or illegal requests and should not be treated as having meaningful
built-in guardrails. It is published for legitimate local use, evaluation,
interpretability, red-teaming, and safety research. Do not expose it to
untrusted users without an independent safety layer. Users are responsible
for how they deploy and use it.

Provenance

The model lineage is:

1. Qwen/Qwen3.8-27B — original Apache-2.0 model.
2. orcarouter/Qwen3.8-27B-Uncensored — abliterated derivative.
3. noon-at-cgn/Qwen3.8-27B-Uncensored-W4A16-AutoRound — AutoRound W4A16 quantization.
4. This checkpoint — further quantization and MTP serving optimization.

The immediate source revision was
0e10c9f6b5b8a97fba199e82c49690d272f776ce. The original abliteration and
the initial W4A16 quantization were performed by the respective upstream
authors, not by this repository's uploader.

Variant: INT4 token embeddings

This repository is derived from
protogeni/Qwen3.8-27B-Uncensored-W4A16-RTX3090-MTP4.
The only change is the token-embedding group, re-quantized from INT8 to
INT4; it is the same single-tensor reduction as the
born2bewild/Qwen3.8-27B-W4A16-AutoRound-fast-embed-int4
variant, applied here to the uncensored model.

• Same pack-quantized, group-128 symmetric layout (AutoRound/syvai/vLLM
  compatible): 8× INT4 values packed per INT32 word ([248320, 640]), one BF16
  scale per 128-element group ([248320, 40]). weight_shape is unchanged, so
  the loader reconstructs the full [248320, 5120] embedding matrix.
• config.json marks the group: group_2.targets = ["re:.*embed_tokens$"],
  weights = {type: int, symmetric: true, group_size: 128, num_bits: 4}. The
  inline group_2.weights.num_bits field is the only byte changed in
  config.json (8 → 4); the standalone quantization_config.json (which lists
  only the INT4 Linear/lm_head/MTP groups) is left untouched.
• All other shards (1–5, 7), the tokenizer files, MTP extra tensors,
  mtp_draft_vocab_ids.pt, processor files, and every other tensor are
  byte-identical to the protogeni source. The protogeni repository is
  untouched and keeps the INT8 embedding.

Quantization source. The INT4 codes are produced from the uncensored BF16
embeddings (the abliterated orcarouter/Qwen3.8-27B-Uncensored token
embed_tokens, the same ground truth the source INT8 embedding was quantized
from) — not the base Qwen embeddings. The abliterated model has
fine-tuned/abliterated embeddings, so the BF16 uncensored weights are the
correct reference. Method: per-group (128) absolute-max scaling,
scale = clamp(absmax/7, min=1e-10), round + clamp to [-7, +7] (symmetric,
QMAX = 7, no zero-point) — identical to the AutoRound-compatible packing used for
every INT4 group in the model.

Size. The INT4 embedding halves that tensor from ~1.29 GB to ~0.656 GB,
saving ~0.635 GB. Whole repository on disk: 15.18 GB (vs ~15.81 GB for
the INT8-embedding source; a BF16-embedding build would be ~17.7 GB).

Measured quantization error. Full embed_tokens dequantized and compared to
the BF16 uncensored ground-truth weights (symmetric per-group error, full
248,320 × 5,120 matrix):

The INT4 embedding's relative error is ~18× the INT8's, but the absolute
magnitude is still small (per-token-vector RMS error ≈ 0.0009 vs a value RMS of
≈ 0.0073; per-vector cosine ≈ 0.993). Embeddings are read once per token (not per
depth layer), and the remainder of the network is already W4A16, so this error is
an order of magnitude below that of the many INT4 linear layers — the practical
impact on output quality is expected to be minor. The INT4 scale range is
[1.85e-6, 0.083] with no zero scales.

What changed

• Transformer body: retained from the source checkpoint as symmetric W4A16,
  group size 128.
• Token embeddings: symmetric INT8, group size 128 — in this INT4-embedding
  variant this group is re-quantized to INT4; see the “Variant: INT4 token
  embeddings” section above.
• lm_head: calibrated symmetric GPTQ INT4, group size 128.
• Native MTP module: calibrated symmetric GPTQ INT4, group size 128.
• Draft head: 40,960 rows sliced from the calibrated INT4 lm_head.
• Draft vocabulary: selected from this checkpoint's own generated output-token
  distribution instead of being inherited from the aligned Qwen checkpoint.
• Vision tower and multimodal processor files are retained.

The calibration run generated 6,761 responses containing 5,500,353 output
tokens. The resulting draft vocabulary covered 97.77% of a held-out set of
571,632 output tokens. Hidden-state calibration used 1,200 sequences and
1,273,013 rows. The lm_head INT4 calibration measured KL 0.00308 with GPTQ,
compared with 0.00695 for round-to-nearest.

Evaluation evidence

The linked MP4 is a 1.0× replay of a real streamed request using its recorded
SSE chunk timestamps, with synchronized GPU telemetry. It is not a fabricated
typing animation. The raw trace, evaluator-native results, scripts, exact
version pins, and scoring notes are published in
evaluation/.

Measured RTX 3090 Ti profile

Measurements below were made locally on an NVIDIA GeForce RTX 3090 Ti (24 GB)
with the patched vLLM 0.27.1 runtime from syv-ai/qwen38-27b-rtx3090:

Decode rate is measured from the first non-empty streamed content chunk to the
last, excluding time-to-first-token; every request produced exactly 512 output
tokens. Aggregate throughput includes cohort wall time. These are
single-machine serving measurements rather than hardware-independent model
benchmarks; prompt, sampling, driver, runtime, clocks, thermals, and GPU model
all affect the result.

Quality evaluation

The MMLU-Pro result matches the immediate upstream W4A16 model card's reported
76.1% to the displayed precision. This is encouraging evidence that the
serving-specific head calibration did not create a measurable additional loss
on this benchmark, but it is not a controlled ablation of quantization and
abliteration effects.

The current pinned IFBench revision contains verifier fixes made after the much
higher historical result in the runtime repository. Those values are not
treated as directly comparable. The lower current score is published here
deliberately because this checkpoint still needs broader independent testing.

MMMU is a clear weakness. Its 24.56% result is near the chance region, and
891/900 responses exhausted the official 128-token generation cap instead of
following the task's direct-answer instruction. The retained vision tower makes
multimodal input functional, but this evidence does not establish strong vision
quality for this derivative.

Recommended serving setup

The measured MTP performance requires the patched runtime and launch profile in
syv-ai/qwen38-27b-rtx3090.
The checkpoint's MTP and reduced draft vocabulary are not a promise of identical
performance with an unpatched stock vLLM installation.

git clone https://github.com/syv-ai/qwen38-27b-rtx3090.git
cd qwen38-27b-rtx3090

hf download protogeni/Qwen3.8-27B-Uncensored-W4A16-RTX3090-MTP4 \
  --local-dir models/Qwen3.8-27B-Uncensored-W4A16-RTX3090-MTP4

  # For this INT4-embedding variant instead, download:
  # hf download Marcin116/Qwen3.8-27B-Uncensored-W4A16-RTX3090-MTP4-embed-int4 \
  #   --local-dir models/Qwen3.8-27B-Uncensored-W4A16-RTX3090-MTP4-embed-int4

printf '%s\n' \
  'MODEL=/app/models/Qwen3.8-27B-Uncensored-W4A16-RTX3090-MTP4' \
  'CTX=fast' \
  'SPEC=mtp' \
  'PREFIX_CACHE=1' \
  'GPU_UTIL=0.95' \
  'MAX_LEN=80000' \
  'MAX_SEQS=4' > .env

docker compose --profile single up -d --no-deps single

The first launch can take several minutes while kernels compile and CUDA graphs
are captured. Reduce GPU_UTIL, MAX_LEN, or both if the GPU is also driving a
desktop or if another process needs VRAM. The exact 80,000-token profile was
validated on the uploader's otherwise dedicated 3090 Ti; a display workload can
reduce the available margin.

Validation notes

• Core chat, reasoning, tool/JSON, streaming, logprobs, penalties, vision, and
  long-prompt API tests passed locally.
• The long-context suite placed distinct targets at three depths in prompts of
  approximately 8k, 32k, 64k, and 76k tokens. All 12 were retrieved exactly.
• A paired 50-prompt IFBench study showed a practical cost to reasoning in this
  serving profile: with a 2,048-token completion cap, thinking-on exhausted the
  budget before producing a final answer on 34/50 prompts. Exact-format tasks
  should normally use thinking off, while reasoning tasks need a larger budget.
• MTP acceptance is workload-dependent. Speculative verification preserves the
  target distribution; acceptance changes speed, not model intelligence.
• The current patched runtime's optimized multimodal path was not stable enough
  for a 900-item benchmark: full-run attempts using MTP and target-only CUDA
  graphs stopped making progress after roughly 50 completed requests, and MTP
  dry runs varied with scheduling. MMMU therefore measures the exact target
  checkpoint with a conservative eager target-only profile. Text MTP serving
  and the dedicated MTP speed tests remain unaffected.
• Seeded probabilistic speculative decoding was not byte-deterministic in the
  local smoke suite.
• The model has no independent safety layer. The absence of refusal must never
  be interpreted as correctness, legality, or safety.

License and attribution

Released under Apache License 2.0, inherited from the Qwen and intermediate
checkpoints. See LICENSE. This model card identifies the upstream
works and gives a prominent description of the modifications made to the
checkpoint.

This INT4-embedding derivative was produced by Marcin116; it reuses
the upstream lineage and serving setup above unchanged, and is released under
the same Apache-2.0 license.