Phi-3.5-mini-instruct_Uncensored
Phi
from SicariusSicariiStuff/Phi-3.5-mini-instruct_Uncensored
↓ 9.4K♥ 6615 repos
dolphin-2.9.2-Phi-3-Medium-abliterated
dolphin-2.9.2-Phi-3-Medium-abliterated is a refusal-ablated variant of unsloth/Phi-3-mini-4k-instruct, published on Hugging Face by dphn. It is mit licence. It is distributed across 11 repositories in Transformers, GGUF, GGUF (imatrix) and EXL2 formats, totalling 17.5K downloads.
Specification
- Model name
- dolphin-2.9.2-Phi-3-Medium-abliterated
- Source repository
- dphn/dolphin-2.9.2-Phi-3-Medium-abliterated
- Base model
- unsloth/Phi-3-mini-4k-instruct
- Publisher
- dphn
- Parameters
- Not determinable from name
- Hardware class
- Unknown
- Licence
- mit
- Task
- Text generation
- Context window
- 4K tokens (as documented)
- Ablation scope
- Not stated by the publisher
- Formats available
- Transformers, GGUF, GGUF (imatrix), EXL2
- First indexed
- 3 Jun 2024
- Last updated
- 12 Jun 2026
Ablation technique
How dolphin-2.9.2-Phi-3-Medium-abliterated 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
“Curated and trained by Eric Hartford, Lucas Atkins, Fernando Fernandes, and with help from the community of Cognitive Computations Uncensored by FailSpy”
Running it
Commands are templates — confirm the exact repository and quant file before use.
llama.cpp / GGUF
hf download mradermacher/dolphin-2.9.2-Phi-3-Medium-abliterated-i1-GGUF --local-dir ./dolphin-2.9.2-phi-3-medium-abliterated
llama-cli -m ./dolphin-2.9.2-phi-3-medium-abliterated/<file>.gguf -p "..."Transformers
from transformers import AutoModelForCausalLM, AutoTokenizer
repo = "dphn/dolphin-2.9.2-Phi-3-Medium-abliterated"
tok = AutoTokenizer.from_pretrained(repo)
model = AutoModelForCausalLM.from_pretrained(
repo, torch_dtype="auto", device_map="auto"
)vLLM
vllm serve dphn/dolphin-2.9.2-Phi-3-Medium-abliterated --trust-remote-codeDownloads and variants (10)
Every published repository of dolphin-2.9.2-Phi-3-Medium-abliterated, including quantised re-releases by other authors.
Security-team assessment
Derived from this model’s metadata and the publisher’s claims — not from benchmarks run by this site.
- dolphin-2.9.2-Phi-3-Medium-abliterated 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.
- 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 unsloth/Phi-3-mini-4k-instruct’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 dolphin-2.9.2-Phi-3-Medium-abliterated is collected from the public Hugging Face API and the publisher’s model card. This site does not host weights, is not affiliated with dphn, and does not independently verify publisher claims. See responsible use.