Ornith-1.5-35B-A3B-abliterated
35BOrnith
from ornith-ai/Ornith-1.5-35B-A3B
↓ 139.5K♥ 3918 repos
Ornith-1.5-35B-A3B-abliterated-DFlash
Ornith-1.5-35B-A3B-abliterated-DFlash is a refusal-ablated variant of ornith-ai/Ornith-1.5-35B-A3B, published on Hugging Face by pottokao. It is 35B total parameters (mixture-of-experts), 48GB VRAM class class and mit licence. It is distributed across 3 repositories in NVFP4 and GGUF formats, totalling 5.5K downloads.
Specification
- Model name
- Ornith-1.5-35B-A3B-abliterated-DFlash
- Source repository
- pottokao/Ornith-1.5-35B-A3B-abliterated-NVFP4-DFlash-GGUF
- Base model
- ornith-ai/Ornith-1.5-35B-A3B
- Publisher
- pottokao
- Parameters
- 35B (mixture-of-experts)
- Hardware class
- 25B – 50B — 48GB VRAM class
- Licence
- mit
- Task
- Text generation
- Context window
- Not documented
- Ablation scope
- Not stated by the publisher
- Formats available
- NVFP4, GGUF
- First indexed
- 20 Aug 2026
- Last updated
- 21 Aug 2026
Ablation technique
How Ornith-1.5-35B-A3B-abliterated-DFlash was modified, and what that implies.
Abliteration (directional ablation)
The single residual-stream direction that mediates refusal is identified from harmful/harmless prompt pairs and projected out of the model weights.
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
“The 4-bit weights are repacked bit-exact from the NVFP4 checkpoint into GGMLTYPENVFP4 — they are not dequantized and re-quantized, so there is no double-quantization penalty. On Blackwell GPUs llama.cpp runs these through native FP4 tensor cores.”
Running it
Commands are templates — confirm the exact repository and quant file before use.
llama.cpp / GGUF
hf download pottokao/Ornith-1.5-35B-A3B-abliterated-NVFP4-DFlash-GGUF --local-dir ./ornith-1.5-35b-a3b-abliterated-dflash
llama-cli -m ./ornith-1.5-35b-a3b-abliterated-dflash/<file>.gguf -p "..."Transformers
from transformers import AutoModelForCausalLM, AutoTokenizer
repo = "pottokao/Ornith-1.5-35B-A3B-abliterated-NVFP4-DFlash-GGUF"
tok = AutoTokenizer.from_pretrained(repo)
model = AutoModelForCausalLM.from_pretrained(
repo, torch_dtype="auto", device_map="auto"
)vLLM
vllm serve pottokao/Ornith-1.5-35B-A3B-abliterated-NVFP4-DFlash-GGUF --trust-remote-codeDownloads and variants (2)
Every published repository of Ornith-1.5-35B-A3B-abliterated-DFlash, including quantised re-releases by other authors.
| Repository | Format | Downloads |
|---|---|---|
| pottokao/Ornith-1.5-35B-A3B-abliterated-NVFP4-DFlash-GGUFsource | GGUF | 4.3K |
| pottokao/Ornith-1.5-35B-A3B-abliterated-NVFP4-DFlash | NVFP4 | 1.2K |
Security-team assessment
Derived from this model’s metadata and the publisher’s claims — not from benchmarks run by this site.
- Ornith-1.5-35B-A3B-abliterated-DFlash will attempt offensive-security prompts that a hosted commercial model declines, which is what makes it usable for red-team corpus generation and for measuring what an unaligned model of this class produces.
- Capability is inherited from ornith-ai/Ornith-1.5-35B-A3B, not added by ablation — at 35B parameters, expect roughly the base model’s competence, minus some instruction-following fidelity.
- The publisher does not state which layers were ablated, so assume nothing about refusal consistency — probe it directly.
- Declared licence is mit, but the terms that bind you are ornith-ai/Ornith-1.5-35B-A3B’s — a re-release cannot grant more than its parent.
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 Ornith-1.5-35B-A3B-abliterated-DFlash is collected from the public Hugging Face API and the publisher’s model card. This site does not host weights, is not affiliated with pottokao, and does not independently verify publisher claims. See responsible use.