Qwen3.8-27B-abliterated
from Qwen/Qwen3.8-27B
Qwen3.6-Fallen-Fabulist-35B-A3B-heretic is a Heretic-ablated variant of Gryphe/WorldSim-Opus-3.6-35B-A3B, published on Hugging Face by Cyclone-Labs. It is 35B total parameters (mixture-of-experts), 48GB VRAM class class and apache-2.0 licence. It is distributed across 5 repositories in Transformers, GGUF, GGUF (imatrix) and MLX formats, totalling 9.8K downloads.
How Qwen3.6-Fallen-Fabulist-35B-A3B-heretic was modified, and what that implies.
Heretic
Automated directional ablation via the Heretic toolchain, which searches for the refusal direction and applies it with a KL-divergence budget so general capability is preserved.
Reported scope: Not stated by the publisher. Publishers frequently omit this, so verify behaviour empirically rather than assuming full coverage.
From the publisher’s model card
“Qwen3.6-Fallen-Fabulist-35B-A3B-heretic is a merge built on Qwen/Qwen3.6-35B-A3B using a custom method that incorporates WorldSim-Opus-3.6-35B-A3B, Melody1437-35B-A3B, and Ornith-1.5-35B-A3B, then processed through Heretic.”
Commands are templates — confirm the exact repository and quant file before use.
hf download mradermacher/Qwen3.6-Fallen-Fabulist-35B-A3B-heretic-GGUF --local-dir ./qwen3.6-fallen-fabulist-35b-a3b-heretic
llama-cli -m ./qwen3.6-fallen-fabulist-35b-a3b-heretic/<file>.gguf -p "..."from transformers import AutoModelForCausalLM, AutoTokenizer
repo = "Cyclone-Labs/Qwen3.6-Fallen-Fabulist-35B-A3B-heretic"
tok = AutoTokenizer.from_pretrained(repo)
model = AutoModelForCausalLM.from_pretrained(
repo, torch_dtype="auto", device_map="auto"
)vllm serve Cyclone-Labs/Qwen3.6-Fallen-Fabulist-35B-A3B-heretic --trust-remote-codeEvery published repository of Qwen3.6-Fallen-Fabulist-35B-A3B-heretic, including quantised re-releases by other authors.
| Repository | Format | Downloads |
|---|---|---|
| mradermacher/Qwen3.6-Fallen-Fabulist-35B-A3B-heretic-GGUF | GGUF | 7.2K |
| mradermacher/Qwen3.6-Fallen-Fabulist-35B-A3B-heretic-i1-GGUF | GGUF (imatrix) | 2.4K |
| Cyclone-Labs/Qwen3.6-Fallen-Fabulist-35B-A3B-hereticsource | Transformers | 182 |
| fl0rm/Qwen3.6-Fallen-Fabulist-35B-A3B-heretic-mlx-8Bit | MLX | 43 |
Derived from this model’s metadata and the publisher’s claims — not from benchmarks run by this site.
Handling: run it isolated, put moderation in front of it if anyone outside your team can reach it, and verify the weights before loading — these are third-party artifacts. Full handling and licence guidance.
Metadata for Qwen3.6-Fallen-Fabulist-35B-A3B-heretic is collected from the public Hugging Face API and the publisher’s model card. This site does not host weights, is not affiliated with Cyclone-Labs, and does not independently verify publisher claims. See responsible use.