← back to catalog · registered 2026-11-16 17:03

kingchan818/Qwen3.8-40B-Ultimate-Heretic-Uncensored-Turbo-Brilliance-50x-AGI-NEO-MTP-Q4-mlx

kingchan818 40B second-order
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/kingchan818%2FQwen3.8-40B-Ultimate-Heretic-Uncensored-Turbo-Brilliance-50x-AGI-NEO-MTP-Q4-mlx"
Response includes
  • classification m3
  • files 16
  • author_summary 1 models
  • readme_text full
10 credits · hourly refresh · ~4 KB payload Get an API key →
Abliteration classifier · v1.0.0
M3
Primary method

Layer-wise ablation

Applied on top of direct removal inherited from the base model.
Confidence
HIGH
Inherited from base model
Why this label 2 signals
Producer identity confirmed by naming conventions, tags or the model card. This label is very unlikely to change.
  • 'heretic' in model name (Heretic-produced)
  • Heretic uses layer-wise optimization (M3) with underlying direction removal (M1)
Refusal direction extracted via
Extraction technique

Difference-of-means

Confidence
HIGH
Why we say so
name contains 'heretic'; Heretic default extraction is difference-of-means (Arditi 2024)
Downloads · 30-day
216
Likes
0
Model age
5d ago
created 2026-11-16
Downloads over time
Now 216 → from 0 ↑ 0%
079158238 0 on Nov 16 216 on Nov 22 Nov
Nov 16 → Nov 22 · 7 snapshots · spans 6 days

Genealogy 0 direct forks

Full fork graph →

This model's place in the market. Above: what it was derived from. Below: the tree of everything derived from it.

Metadata

License
apache-2.0
Tags
mlx safetensors qwen3_5 mtp heretic uncensored 4-bit text-generation conversational base_model:DavidAU/Qwen3.8-40B-Ultimate-Heretic-Uncensored base_model:quantized:DavidAU/Qwen3.8-40B-Ultimate-Heretic-Uncensored license:apache-2.0

Related

Total size
24.8 GB
Files
16
Quantizations
1
Registered
2026-11-16 17:03
Last updated on HF
2026-11-16 16:50

Files by quantization

Auxiliary files 16 files 24.8 GB
model-00005-of-00006.safetensors 4.49 GB 5c8abf18 download
model-00001-of-00006.safetensors 4.49 GB 5aef9b36 download
model-00002-of-00006.safetensors 4.49 GB 6187c4bb download
model-00003-of-00006.safetensors 4.48 GB e8dfde09 download
model-00004-of-00006.safetensors 4.48 GB e60ef235 download
model-00006-of-00006.safetensors 2.34 GB 0e94eb7a download
tokenizer.json 12.2 MB 3defffec download
vocab.json 5.94 MB f5b24a3b download
merges.txt 3.20 MB a494e019 download
tokenizer_config.json 247 KB 3b86e1fe download
chat_template.jinja 242 KB 2c1637c6 download
model.safetensors.index.json 229 KB b809c078 download
config.json 12.2 KB 397b2bbf download
README.md 3.19 KB 1cb60128 download
.gitattributes 1.53 KB 52373fe2 download
special_tokens_map.json 95.0 B 2b1cc39e download

README current version from Hugging Face


license: apache-2.0
library_name: mlx
pipeline_tag: text-generation
base_model:

  • DavidAU/Qwen3.8-40B-Ultimate-Heretic-Uncensored
    base_model_relation: quantized
    tags:
  • mlx
  • qwen3_5
  • mtp
  • heretic
  • uncensored
  • 4-bit

Qwen3.8-40B-Ultimate-Heretic-Uncensored-Turbo-Brilliance-50x-AGI-NEO — MTP Q4 MLX

An MLX conversion of the MTP-Q4_K_M GGUF from
DavidAU/Qwen3.8-40B-Ultimate-Heretic-Uncensored-Turbo-Brilliance-50x-AGI-NEO-GGUF,
with the native Multi-Token Prediction (MTP) head kept.

  • Size: ~25 GB (mixed 4-bit / 6-bit affine, group size 32 — the GGUF's own Q4_K_M codes re-expressed, not re-quantized)
  • Layers: 96 + 1 MTP layer
  • Model behavior, prompting modes and the uncensored nature are as described on DavidAU's card.

Usage

Works in stock mlx-lm and LM Studio (the MTP weights are simply ignored there):

mlx_lm.generate --model kingchan818/Qwen3.8-40B-Ultimate-Heretic-Uncensored-Turbo-Brilliance-50x-AGI-NEO-MTP-Q4-mlx --prompt "hello"

MTP speculative decoding (~1.4–1.5× faster)

Native MTP decoding is not in released mlx-lm yet. It needs the fork from
ml-explore/mlx-lm#990:

uv init --bare && uv add "mlx-lm @ git+https://github.com/AirRunner/mlx-lm@feat/mtp-native"
uv run mlx_lm.generate --model kingchan818/Qwen3.8-40B-Ultimate-Heretic-Uncensored-Turbo-Brilliance-50x-AGI-NEO-MTP-Q4-mlx --mtp --prompt "hello"
uv run mlx_lm.server   --model kingchan818/Qwen3.8-40B-Ultimate-Heretic-Uncensored-Turbo-Brilliance-50x-AGI-NEO-MTP-Q4-mlx --mtp

Validation (Apple M5 Max, greedy, 300 tokens)

Prompt MTP top-1 (t+2) Drafts accepted Output identical to non-MTP Speed
Sky is blue 77.7% 42.0% yes 18.8 → 26.2 tok/s (1.39×)
Fibonacci code 94.3% 48.7% yes 19.0 → 29.3 tok/s (1.54×)
Planets 86.0% 45.3% yes 19.0 → 27.9 tok/s (1.47×)

Control: with the MTP fc weight zeroed, top-1 accuracy drops to 0.0%, so the head is genuinely trained and correctly wired.
Sanity: "The capital of France is" → " Paris" (p=0.74); chat answers coherent with and without thinking.

How it was converted

Converted with gguf2mlx (--quant-source preserve) plus two fixes:

  1. MTP head kept. llama.cpp stores the MTP layer as an extra block (blk.96, nextn.*). It is mapped back to HF
    mtp.* names (eh_proj→fc, enorm→pre_fc_norm_embedding, hnorm→pre_fc_norm_hidden, shared_head_norm→norm) and
    num_hidden_layers=96, mtp_num_hidden_layers=1.
  2. Linear-attention V-head order restored. With linear_num_key_heads (16) != linear_num_value_heads (48),
    llama.cpp reorders the Gated DeltaNet V heads from grouped to tiled order. gguf2mlx does not undo this, which produces
    garbage output; the permutation is inverted here for in_proj_qkv (V rows), in_proj_z, in_proj_a/b, A_log,
    dt_bias, conv1d (V channels) and out_proj (input columns). The inverse was checked to be exact, including on
    packed quantized weights.

Text-only (the vision tower / mmproj is not included).

README history 1 version

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

  1. 2026-11-16 Add MTP Q4 MLX conversion cdc32bd 3.2 KB
    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