← back to catalog · registered 2026-09-20 06:56

JC1DA/DavidAu-Qwen3.8-27B-TWIN-TURBO-Fable-Cold-Fusion-709-ULTRA-HERETIC-Uncensored-INT4-INT8-Mixed

curl -H "Authorization: Bearer $ABL_KEY" \
     "https://abliteration.org/api/v1/models/JC1DA%2FDavidAu-Qwen3.8-27B-TWIN-TURBO-Fable-Cold-Fusion-709-ULTRA-HERETIC-Uncensored-INT4-INT8-Mixed"
Response includes
  • classification unknown
  • files 20
  • author_summary 4 models
  • readme_text full
10 credits · hourly refresh · ~4 KB payload Get an API key →
Refusal direction extraction

No specific extraction method could be identified for this model. The producer either did not document it or used a proprietary pipeline.

What is a refusal direction? →
Downloads · 30-day
0
Likes
0
Model age
today
created 2026-09-20

Metadata

Tags
safetensors qwen3_5 8-bit compressed-tensors region:us

Related

Total size
23.4 GB
Files
20
Quantizations
1
Registered
2026-09-20 06:56
Last updated on HF
2026-09-20 06:51

Files by quantization

Auxiliary files 20 files 23.4 GB
model-00004-of-00008.safetensors 2.99 GB a7a094a1 download
model-00006-of-00008.safetensors 2.98 GB 65ee68f4 download
model-00001-of-00008.safetensors 2.98 GB d0dc26ec download
model-00005-of-00008.safetensors 2.98 GB e38bfaba download
model-00002-of-00008.safetensors 2.95 GB 097a75de download
model-00003-of-00008.safetensors 2.92 GB 3a9bf26e download
model-00007-of-00008.safetensors 2.41 GB d99d747f download
model-00008-of-00008.safetensors 2.37 GB cd1dd19a download
model_extra_tensors.safetensors 810 MB 9f1f9e40 download
tokenizer.json 19.1 MB 06b95093 download
model.safetensors.index.json 178 KB b1d2b43e download
config.json 45.5 KB 3dc2f120 download
quantization_config.json 40.2 KB 1d7412e9 download
chat_template.jinja 16.7 KB 6e25105d download
README.md 3.34 KB fbfc54da download
.gitattributes 1.53 KB 52373fe2 download
processor_config.json 1.16 KB 33818c7f download
tokenizer_config.json 1.14 KB 1d134cd2 download
preprocessor_config.json 390 B 2ea84a43 download
generation_config.json 213 B 62c03847 download

README current version from Hugging Face

Qwen3.8-27B-TWIN-TURBO-Fable-Cold-Fusion-709-ULTRA-HERETIC-Uncensored (W8A8 Mixed)

Mixed-precision quantized version of Qwen3.8-27B-TWIN-TURBO-Fable-Cold-Fusion-709-ULTRA-HERETIC-Uncensored.

Quantization

Aspect Detail
Method compressed-tensors (Unsloth 2026.7.2)
Format mixed-precision
Source precision bfloat16 (~55.6 GB, 12 shards)
Quantized size ~25.1 GB (8 shards + 0.85 GB extra tensors)
Size reduction ~55%

Precision groups

Group Weight bits Activation bits Strategy Layers Format
W8A16 8 (int, symmetric) BF16 (unquantized) channel, static 66 pack-quantized
W8A8 8 (int, symmetric) 8 (int, symmetric) channel + token (dynamic) 143 int-quantized
W4A16 4 (int, symmetric) BF16 (unquantized) group (size 128), static 191 pack-quantized

Layers kept at full BF16 precision (excluded from quantization): all 26 vision blocks, all linear-attention projection layers (in_proj_a, in_proj_b, norm, linear_attn), and lm_head. These are stored in model_extra_tensors.safetensors.

Architecture

  • Model type: qwen3_5 (Qwen3_5ForConditionalGeneration — multimodal, text + vision)
  • Hidden size: 5120
  • Layers: 64 (linear-attention with every 4th layer being full attention)
  • Attention heads: 24 Q / 4 KV (full); 16 key / 48 value (linear)
  • Head dim: 256 (full) / 128 (linear)
  • Context: 262 144 tokens
  • Vocab size: 248 320
  • Parameters: ~13.4 B (post-quantization weight storage: 24.25 GB)

Benchmark

AIME 2026 (both AIME I and II)

Metric W8A8 (this repo) FP16 baseline
Accuracy 90.0% (27/30) 86.7% (26/30)
Problems 30 (15 AIME I + 15 AIME II) 30 (15 AIME I + 15 AIME II)
Protocol 0-shot, boxed numeric answer 0-shot, boxed numeric answer
Avg latency 216.7 s 356.3 s
Avg output tokens 16 625 15 590
Total tokens (run) 31 003 input / 498 739 output 31 003 input / 467 694 output
Parallelism 8 concurrent requests 8 concurrent requests

Run date: 2026-09-19 (both runs).

  • W8A8 (this repo): one problem hit the 100K max-tokens cap (likely a false negative from truncated reasoning).
  • FP16 baseline: 4 problems produced no \boxed{} answer (reasoning ran to the token cap without a final answer); the remaining 26 were correct. Within run-to-run noise on a 30-problem set, the quantized W8A8 scores on par with (slightly above) full FP16.

Usage

Load with SGLang or vLLM using the quantization config in this directory:

# SGLang
sglang.launch_server --model-path /path/to/this-dir --port 8002

# vLLM (if supported by your vLLM version)
vllm serve /path/to/this-dir --port 8002

The quantization_config.json is automatically detected by runtimes that support the compressed-tensors format. No additional flags are needed for the mixed-precision scheme.

Source model

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 app" button that hands off directly to a local runtime of your choice - Infrahuman, LM Studio, or Ollama. No API keys, no subscription, no prompt leakage.