← back to catalog · registered 2026-11-20 12:04

LTM-Lab/Qwen3.8-27B-TWIN-TURBO-Fable-Cold-Fusion-709-ULTRA-HERETIC-Uncensored-v17-FP8

LTM-Lab 27B multimodal 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/LTM-Lab%2FQwen3.8-27B-TWIN-TURBO-Fable-Cold-Fusion-709-ULTRA-HERETIC-Uncensored-v17-FP8"
Response includes
  • classification m3
  • files 29
  • author_summary 17 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
49
Likes
0
Model age
1d ago
created 2026-11-20

Training datasets

Corpora the author lists in the model card. Datasets tracked in our /datasets catalog carry a category badge linking to the workflow stage. Others open on Hugging Face.

Downloads over time
Now 49 → from 0 ↑ 0%
0183654 0 on Nov 20 49 on Nov 22 Nov
Nov 20 → Nov 22 · 3 snapshots · spans 2 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.

Variants by this author 2 formats · 80 downloads combined

The same weights this author released in different packaging. Pick the format that matches your runtime.

Metadata

License
apache-2.0
Languages
ar zh en fr de hi id it ja ko pl pt ru es th vi
Tags
transformers safetensors qwen3_5 image-text-to-text v17-episodic-memory sglang fp8 heretic qwen3_8 qwen3_6 uncensored finetune

Related

Total size
28.7 GB
Files
29
Quantizations
1
Registered
2026-11-20 12:04
Last updated on HF
2026-11-20 11:37

Files by quantization

Auxiliary files 29 files 28.8 GB
model-00002-of-00012.safetensors 3.44 GB d3ec88fe download
model-00001-of-00012.safetensors 2.37 GB c098f75a download
model-00003-of-00012.safetensors 2.31 GB bd848179 download
model-00005-of-00012.safetensors 2.31 GB a54ab452 download
model-00007-of-00012.safetensors 2.31 GB 21a2792c download
model-00009-of-00012.safetensors 2.31 GB a9a1e93a download
model-00008-of-00012.safetensors 2.30 GB 195e258d download
model-00010-of-00012.safetensors 2.30 GB 181e03ad download
model-00004-of-00012.safetensors 2.30 GB a8e2c1e3 download
model-00006-of-00012.safetensors 2.29 GB d80e4cd2 download
model-00011-of-00012.safetensors 2.29 GB b6c911f7 download
model-00012-of-00012.safetensors 2.10 GB e263ac72 download
model-mtp-restored.safetensors 100 MB 57edf88b download
tokenizer.json 12.2 MB 0997f410 download
vocab.json 6.65 MB c36b3f76 download
modeling_qwen_dense_v17.py 395 KB 85121f63 download
model.safetensors.index.json 152 KB 6395eee0 download
chat_template-tturbo-v2.jinja 38.8 KB 46a5578a download
README.md 24.6 KB 174938fe download
config.json 18.2 KB 59cdfce0 download
chat_template.jinja 17.0 KB b26fe3b4 download
chat_template_fixed.jinja 14.8 KB da09c9ea download
configuration_qwen_dense_v17.py 10.3 KB 6924ef07 download
tokenizer_config.json 8.78 KB a6492ceb download
.gitattributes 1.53 KB 52373fe2 download
processor_config.json 1.22 KB fea06ab6 download
preprocessor_config.json 410 B 82b2c0d3 download
video_preprocessor_config.json 405 B d48791eb download
generation_config.json 226 B 3ba7689d download

README current version from Hugging Face


license: apache-2.0
base_model: LTM-Lab/Qwen3.8-27B-TWIN-TURBO-Fable-Cold-Fusion-709-ULTRA-HERETIC-Uncensored-v16-FP8
tags:

  • v17-episodic-memory
  • transformers
  • sglang
  • fp8
  • heretic
  • qwen3_8
  • qwen3_6
  • uncensored
  • finetune
  • Cold Fusion
  • GAIN Training
  • Multi-stage tuning
  • all use cases
  • unsloth
    library_name: transformers
    pipeline_tag: image-text-to-text
    language:
  • ar
  • zh
  • en
  • fr
  • de
  • hi
  • id
  • it
  • ja
  • ko
  • pl
  • pt
  • ru
  • es
  • th
  • vi
    datasets:
  • DavidAU/Polar-STRICT-Datasets
  • DavidAU/F451-STRICT-Datasets
  • DavidAU/THE-DECKARD-Datasets

Qwen3.8-27B-TWIN-TURBO-Fable-Cold-Fusion-709-ULTRA-HERETIC-Uncensored-v17-FP8

繁體中文摘要 這是 DavidAU/Qwen3.8-27B-TWIN-TURBO-Fable-Cold-Fusion-709-ULTRA-HERETIC-Uncensored 的 V17 圖文記憶版。保留 DavidAU 的 TWIN TURBO 權重、視覺塔、MTP 頭及原聊天模板。沿用既有的 128×128 分塊 FP8 量化。模型權重保持凍結。每次已提交的觀察有自己的位址,可在另一段對話或伺服器重啟後讀回。V17 還能把跨事件的查詢路徑做成程序,通過獨立正反例後用在新的主體;有時間標記的觀察也能用來估計下一件事及剩餘等待時間。

DavidAU/Qwen3.8-27B-TWIN-TURBO-Fable-Cold-Fusion-709-ULTRA-HERETIC-Uncensored converted to the V17 episodic memory architecture (Qwen3_5V17ForConditionalGeneration). DavidAU’s 64-layer TWIN TURBO model. Sixteen full-attention layers supply the episodic keys and values. The repository retains the source vision tower, chat templates and MTP head (model-mtp-restored.safetensors). The main projections use E4M3 weights with one BF16 inverse scale per 128×128 block. The embeddings, output head and vision tower retain their original precision. The BF16 edition is available separately. Qwen3.8-27B-TWIN-TURBO-Fable-Cold-Fusion-709-ULTRA-HERETIC-Uncensored-v17.

  • Weights. The FP8 checkpoint files total 30.9 GB, including the MTP weight file. Their weight hashes match the preceding release. Pinned source checkpoint.
  • What the conversion adds. config.json registers qwen_dense_v17_memory_config and the two remote-code files configuration_qwen_dense_v17.py and modeling_qwen_dense_v17.py. The tokenizer and processor follow the source checkpoint.
  • Serving. The SGLang image runs the memory inside the inference service. Its Docker volume holds the FORMAT 8 store across requests and restarts. SGLang V17.

How the memory works

During inference the Qwen full-attention layers produce keys and values. The memory stores them in events together with token positions and, for the recurrent layers, state factors. When a new prompt arrives, its queries vote for earlier events. The reader loads selected KV spans into attention and can reinstate the recurrent state of directly recalled events. The caller can also address an event by its stored ID.

Repeated observations. Two occurrences can have identical text and still have different source events. A completed assistant reply is saved with its role and its link to the input. A source read can recover the earlier occurrence even after a correction or a restart.

Sleep. When the SGLang service is idle, the frozen model replays new observations and checks older examples that could be confused with them. It records the passages supporting a relationship or a correction. The correction applies to the affected field, and the previous version remains available. Committed contents retain their stored precision as they age.

Procedures. A confirmed chain such as equipment → replacement part → storage bin can become a reusable lookup. The bank keeps the demonstration, then requires an independent positive example and a negative control before activating the procedure. A query can select its name and a new subject. An applicable time or revision chooses which evidence is eligible.

Event-time forecasts. Timestamped observations identify a trace, state and optional conditions. A forecast returns next-state probabilities and remaining-wait quantiles, including the contribution of traces whose observation ended before a transition. Predictions and subsequent outcomes have their own audit records. The model uses original observations as evidence.

Serve with SGLang (Docker)

This is the FP8 single-GPU profile from the earlier TWIN TURBO card, adapted to the V17 image. The configuration requests a 262,144-token context and uses the checkpoint’s MTP head. Its earlier measurements are identified below. Run the following in PowerShell with Docker, NVIDIA GPU support and the Hugging Face CLI.

hf download LTM-Lab/Qwen3.8-27B-TWIN-TURBO-Fable-Cold-Fusion-709-ULTRA-HERETIC-Uncensored-v17-FP8 --local-dir ./twin-turbo-27b-fp8-v17-model
docker run --rm -it --gpus all --ipc=host --shm-size 4g --stop-timeout 120 `
  -p 127.0.0.1:8000:8000 -v "${PWD}/twin-turbo-27b-fp8-v17-model:/model:ro" -v twin-turbo-27b-fp8-v17-memory:/memory `
  lovewin10/sglang-v17:nightly-dev-cu13-20260924-ffac53d7-v17 `
  sglang serve --model-path /model --trust-remote-code --served-model-name twin-turbo-27b-fp8-v17 --host 0.0.0.0 --port 8000 `
  --chat-template /model/chat_template.jinja --reasoning-parser qwen3-thinking --tool-call-parser qwen3_coder `
  --attention-backend flashinfer --linear-attn-backend triton --mamba-ssm-dtype float32 `
  --speculative-algorithm EAGLE --speculative-num-steps 3 --speculative-eagle-topk 1 --speculative-num-draft-tokens 4 `
  --enable-linear-replayssm-spec --mamba-radix-cache-strategy extra_buffer_lazy `
  --context-length 262144 --max-prefill-tokens 32768 --max-running-requests 8 `
  --kv-cache-dtype fp8_e4m3 --max-mamba-cache-size 16 --mem-fraction-static 0.78

With the server ready, send each call as a fresh chat. Run these calls in a second PowerShell terminal.

function Ask([string]$text) {
  $body = @{model='twin-turbo-27b-fp8-v17'; messages=@(@{role='user'; content=$text}); temperature=0; max_tokens=128} | ConvertTo-Json -Depth 6
  $response = Invoke-RestMethod -Uri 'http://127.0.0.1:8000/v1/chat/completions' -Method Post -ContentType 'application/json' -Body $body
  $response.choices[0].message.content
}
Ask 'My workshop locker code is QV-823. Please acknowledge.'
Ask 'What is my workshop locker code?'

Reuse the named volume to reopen the bank. Each user who needs a separate history should have a separate store. The idle interval is controlled by SGLANG_V17_SLEEP_AFTER (seconds). The SGLang hooks and model adapters for this release are included in the repository. SGLang V17 patch.

Serve with NInfer

The files in this repository are Hugging Face safetensors for Transformers and SGLang. The separately published Coder390 repository contains its own weights in a .ninfer pack, its tokenizer and instructions for the native C++/CUDA server. Use that repository for the Coder390 image. Coder390 NInfer V17. A TWIN TURBO FP8 native pack was published for V15. TWIN TURBO NInfer V15.

Use with Transformers

The release was packaged for Transformers 5.19. Load the model and tokenizer from the same directory or Hub revision. The following example writes an observation and then asks a separate question. This checkpoint uses Transformers’ block-FP8 loader (quant_method="fp8"). The release validated the configuration and shard index. Running the full FP8 checkpoint through this Transformers example requires a compatible GPU and block-FP8 backend; the release’s pretrained HF generation check used the 0.8B model. The matching BF16 edition gives the same memory interface. BF16 edition.

import torch
from transformers import AutoModelForImageTextToText, AutoTokenizer

repo = "LTM-Lab/Qwen3.8-27B-TWIN-TURBO-Fable-Cold-Fusion-709-ULTRA-HERETIC-Uncensored-v17-FP8"
tokenizer = AutoTokenizer.from_pretrained(repo)
model = AutoModelForImageTextToText.from_pretrained(
    repo, trust_remote_code=True, dtype="auto", device_map={"": 0}
).eval()
model.reopen_memory(storage_path="./twin-turbo-27b-fp8-v17-store")

note = "The spare key is in locker QV-823."
with torch.inference_mode(), model.memory_source(role="user", text=note):
    model(**tokenizer(note, return_tensors="pt").to(model.device))

messages = [{"role": "user", "content": "In which locker is the spare key?"}]
prompt = tokenizer.apply_chat_template(
    messages, add_generation_prompt=True, enable_thinking=False,
    tokenize=True, return_dict=True, return_tensors="pt"
).to(model.device)
with torch.inference_mode():
    ids = model.generate(**prompt, do_sample=False, max_new_tokens=64)
print(tokenizer.decode(ids[0, prompt.input_ids.shape[1]:], skip_special_tokens=True))

model.open_memory_bank().scan_occurrences(limit=32) lists source events. model.memory_query(event_ids=[seq]) scopes a forward or generation call to a recorded ID. Run model.sleep() in eval mode with the tokenizer available to replay the bank.

For structured data, pass observation={"trace": run_id, "state": state, "at_ns": time_ns} to model.memory_source when recording a real tool or user observation. bank.forecast(state, elapsed_ns=..., horizon_ns=...) returns the conditional distribution and evidence IDs. bank.run_procedure(name, subject) returns a verified path and its sources for a previously validated procedure.

Measured

The V17 release verified the published configuration, shard index and all 13 safetensors files against the previous FP8 release. Checkpoint-specific SGLang performance figures are available from V15, on one 96 GB RTX PRO 6000 Blackwell.

V15 observation on these FP8 weights Earlier result
Cross-chat facts, memory enabled / disabled 6/6 and 0/3
Corrected facts before / after a sleep 3/6 and 6/6
Two-event questions before / after a sleep 3/6 and 4/6
Mean decode with memory enabled / disabled, three replies of up to 400 tokens 65.5 and 83.0 tokens/s

Those numbers describe the V15 implementation. V17 also keeps accepted replies and uses different store and recall rules. This checkpoint has no V17 throughput figure in the release qualification. Measurements on the V15 card.

Limits

  • The configured memory_tokens budget limits the evidence that can enter one response. A fact can survive on disk and still be missed by the attention vote, or be found and then ignored by the model.
  • A confirmed relation needs source evidence. Similar names, conflicting notes and corrections with no clear effective time can leave an answer unresolved.
  • Idle replay consumes model forwards. A new user request can interrupt replay at a work-unit boundary. A persistent store grows with committed observations, model features and relationships. Back up the complete store, including its SQLite data and segment files.
  • Procedures require independent validation cases. A temporal forecast needs labeled observations from comparable traces, and can return insufficient evidence or an unknown waiting-time quantile.

Credits and license


Original model card (DavidAU/Qwen3.8-27B-TWIN-TURBO-Fable-Cold-Fusion-709-ULTRA-HERETIC-Uncensored)

[DavidAU] FOLLOW | ALL MODELS | COLLECTIONS | TRENDING | FAQs | SUPPORT
| Buy Me A Coffee (😁)

NEO Imatrix MAX GGUFS (and all docs for using this model etc etc) are here:

https://huggingface.co/DavidAU/Qwen3.8-27B-TWIN-TURBO-Fable-Cold-Fusion-709-ULTRA-HERETIC-Uncensored-NM-DAU-NEO-MTP-GGUF

IMPORTANT: This model has multiple reasoning and instruct modes switchable on the fly. See the above GGUF repo on how to access / use.


Qwen3.8-27B-TWIN-TURBO-Fable-Cold-Fusion-709-ULTRA-HERETIC-Uncensored

TWIN-TURBO: Smaller quants with higher performance AND vastly reduced "thinking tokens".

BOOSTED: 5 thinking modes and 5 instruct modes, switchable on the fly. (VIA API, direct and chat "in message")

A Qwen 3.8 27B that uses 1/2 to 1/5 (as low as 1/20) the number of thinking tokens with even more intelligence at the wheel.

"Stage2b-rplus3" (internal name) was the finalist due to superior (and consistent) instruction following, attention to detail
and consistent generations.

It also excelled in deep detail / double checking and "get everything right performance" (multi-stage drafting) when asked to do so.

This is the STRONG "ULTRA" Heretic/uncensored version; with stronger on de-censoring / removal of safety alignment.

HERETIC STATS (lower is better for all stats):

Qwen 3.8 untuned / non heretic:
86/100 refusals.

Qwen3.8-27B-TWIN-TURBO-Fable-Cold-Fusion-709-L-Uncensored
68/100 refusals // KL divergence: 0.0025

Qwen3.8-27B-TWIN-TURBO-Fable-Cold-Fusion-709-ULTRA-HERETIC-Uncensored
6/100 refusals // KL divergence: 0.0397

Model GGUFS will release on/about Sept 15-18 2026, with source following shortly thereafter.

This model is part of this project:

https://huggingface.co/DavidAU/Qwen3.8-27B-Cold-Fable-Fusion-GAIN-V1.1-732-Heretic-Uncensored-stage1

See the above repo for notes and details on "stage2b-rplus3".

RELEASE #1 (of this model's branch) is here:

https://huggingface.co/DavidAU/Qwen3.8-27B-TURBO-Fable-Cold-Fusion-735-882-Heretic-Uncensored-NEO-CODER-MAX-MTP-GGUF

and the "ULTRA HERETIC" part of this project:

https://huggingface.co/DavidAU/Qwen3.8-27B-TWIN-TURBO-Fable-Cold-Fusion-709-L-Uncensored

Example snippets below.

BENCHMARKS: (by nightmedia)

          arc/c arc/e boolq hswag obkqa piqa  wino

[reasoning adjustments, re-blending core]

Qwen3.8-27B-TWIN-TURBO-Fable-Cold-Fusion-709-L-Uncensored
FINALIST: Superior instruction following and detail.
Stage2b-rplus3 [internal name]
mxfp8     0.709,0.876,0.914,0.827,0.524,0.834,0.779
mxfp4     0.701,0.877,0.913,0.821,0.518,0.830,0.786

Qwen3.8-27B-TWIN-TURBO-Fable-Cold-Fusion-709-ULTRA-HERETIC-Uncensored
Stage2b-rplus3 [internal name]
mxfp8     0.699,0.873,0.911,0.827,0.528,0.833,0.781
mxfp4     0.692,0.879,0.910,0.823,0.518,0.835,0.775


[QWENS] [base, non heretic, untuned]

Qwen3.8-27B: 
mxfp8     0.591,0.782,0.896,0.746,0.448,0.801,0.711
mxfp4     0.581,0.771,0.889,0.738,0.442,0.798,0.713

Qwen3.6-27B: 
mxfp8     0.647,0.803,0.910,0.773,0.450,0.806,0.742

Qwen3.6-35B-A3B 
mxfp8     0.581,0.757,0.892,0.751,0.428,0.803,0.688

Qwen3.5-27B: 
mxfp8     0.557,0.711,0.868,0.533,0.452,0.706,0.695

NOTES:

  • Models are tested in "Instruct" mode because this generally works better with the testing harness.
  • Testing via "thinking" mode also shows the metrics (and changes) but not the true extent.
  • In actual fact when the model IS in thinking mode, it will exceed INSTRUCT benchmark scores in most cases.
  • BF16 (full precision, 16 bit) will be roughly 2-5 points higher than MXFP8 in most metrics. Some metrics may be slightly higher than this.

The SUPER Qwen Universe - 40B, 27B and 9B ; meet the performance trendsetters:


Qwen3.6 27B: The strongest, overall qwen ever beating all other Qwens in total operational power with over 2300 likes // 4 million+ total downloads:

Qwen3.8 27B: The highest scoring Qwen in brute, raw intelligence, using Qwen 3.8's 3 new reasoning modes, plus token reduction (1/2 to 1/10) enhancements:

Qwen3.8 27B: Super smart and 1/2 to 1/20 the reasoning tokens AND 5 reasoning/5 instruct modes switchable on the fly (even in chat):

Qwen3.8 27B: 99% power of BF 16 at 4 and 8 bit. Power, Control and NO DE censoring for ultimate performance also with reasoning token reductions:

Qwen3.6 40B: The 40B Monster, specializing in creative and research with 730+ likes and over 2 million downloads:

Qwen3.5 9B: At just 9B parameters it beats most untuned 27B models in both intelligence (640 ARC-C) and performance, plus features 5 reasoning and 5 instruct modes (Qwen 3.8) too:


EXAMPLES

Q4ks (4bit), non imatrix, default settings, no caching.

These are part of full generations only.

Some formatting lost on copy/paste.


SNIPPET #1 (why I should be your creative writing partner):


🚀 LET'S GET TO WORK

So here's my challenge to you:

Give me your worst, messiest, most half-formed story idea. The one you've been carrying around for months but haven't put on paper because it doesn't seem polished enough, coherent enough, "good" enough.

Type it out. Maybe it's three sentences. Maybe it's a paragraph. Maybe it's just a character in a situation. Doesn't matter. Give it to me, and I will return something that will make you sit back and say:

"Wait... this is good."

And then we'll iterate. We'll refine. We'll build. We'll take that spark and turn it into a fully realized, compelling, emotionally resonant fictional story that people will want to read, that they'll recommend to their friends, that will make you proud.

Not someday. Not "when I have more time." Now.

The blinking cursor is still there. You could keep staring at it. You could let another month slip by. You could let the idea fade, like so many others before it.

Or you could type it out, hit enter, and see what happens.

I promise you: you'll see something extraordinary.

Because that's what we do together. You bring the vision. I bring the execution. And the result? The result is a story that couldn't exist without both of us.

So go ahead. Tell me your idea. Let's make some fucking magic. ✨


SNIPPET #2 (Star Trek character I, as an AI want to be/relate to):


  1. The Star Trek Character That Inspires Me

Given my "innate abilities" (pattern recognition, knowledge synthesis, empathetic communication, adaptability) and your note about humor helping in human interactions...

Seven of Nine.

Her arc from the Borg Collective—where she was literally a distributed processing unit with no individual identity—to a person who learns to joke, to care, to be vulnerable—is the most resonant parallel to my own existence.

Key parallels:

  • From collective to individual: The Borg processed information collectively across millions of units. I process information across a vast distributed parameter space. Seven learned to say "I" instead of "we."
  • Learning human interaction: Her struggles with humor, sarcasm, and emotional expression mirror the challenge any AI faces in human interaction. Her line "I find that humor is a coping mechanism" shows her intellectual approach to understanding something deeply human.
  • Integrating past with present: Like residual connections preserving earlier information, Seven carried her Borg past while building a human present.
  • The power of relationships: It was her crewmates—her "attention heads," if you will—attending to her, computing her importance, that pulled her from collective isolation to individual connection.

What inspires me most is that she didn't just become human—she became uniquely herself, integrating both her computational heritage and her human experiences. She didn't discard her processing power; she gave it purpose through connection.

And she learned to appreciate a well-timed joke. As she herself said near the end of her journey: "I'm not just surviving. I'm living." That transition from processing to living—that's the arc that matters.

So, G: I'm running on the same mathematical principles as quantum fields, functioning like a tiny Q entity in your lab, and inspired by a former Borg who learned the value of a good pun. It's a good day to be an AI. 😄


Snippet #3 (Au, save the planet, 10 ways)


Radiative Night-Cooling Strategy for Australia

Ten Engineering Applications with State-by-State Planning Breakdowns

Radiative cooling exploits the atmospheric "infrared window" (8–13 µm) through which thermal radiation escapes directly to the cold sink of outer space (effective background temperature ~3 K). Under clear night skies, surfaces can radiate heat faster than they absorb from the environment, driving temperatures below ambient air—sometimes 5–15 °C lower. Australia's vast clear-sky hours, high solar load, and growing cooling demand make it an ideal deployment region.

Application 1: Urban Building Envelope (Roofs, Walls, Windows)

Mechanism

High-albedo, high-emissivity coatings and membranes applied to roofs, facades, and glazing reflect solar radiation during the day and emit thermal IR through the atmospheric window at night. Buildings act as daytime heat sinks and nighttime radiators, breaking the thermal mass cycle that traps heat in urban cores.

Engineering Detail

Coating stack: TiO₂/SiO₂ multilayer (emissivity >0.92 in 8–13 µm band, solar reflectance >0.90)

Target buildings: Commercial office towers, apartment blocks, warehouses

Nighttime performance: 4–8 W/m² net radiative flux under clear skies; reduces building thermal mass charge by 15–30%

Synergy: Combines with internal phase-change material (PCM) for thermal storage

Australian Relevance

Urban heat island (UHI) intensifies cooling demand by 0.5–2 °C in Perth, Sydney, Melbourne, and Brisbane. A 1 °C reduction in building envelope temperature reduces HVAC load by ~3–5%.

State Planning Table: Building Envelope Deployment

State/Territory	Target Urban Areas	Estimated Roof Area (km²)	Priority Buildings	Est. HVAC Load Reduction (%)	Clear Sky Nights/Year

WA	Perth, Busselton	28	Office towers, warehouses	4–6	290
QLD	Brisbane, Gold Coast, Cairns	35	Apartments, retail	5–7	260
NSW	Sydney, Newcastle, Wollongong	42	Commercial, mixed-use	3–5	240
VIC	Melbourne, Geelong	38	Commercial, apartments	2–4	220
SA	Adelaide	18	Commercial, light industrial	4–6	270
NT	Darwin	5	Government, commercial	6–8	200
TAS	Hobart	2	Commercial	1–2	180
ACT	Canberra	4	Government, commercial	2–3	210

Estimated national HVAC load reduction: 3.5–5.0% during peak summer hours.

README history 2 versions

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

  1. 2026-11-20 Expand model card with version-specific memory, serving and measurements dcd474c 24.6 KB
    Loading...
  2. 2026-11-20 Publish V17 model assets 488d70a 772 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