⌦abliteratedmodels.org

Dolphin3.0-R1-Mistral-24B-abliterated

Dolphin3.0-R1-Mistral-24B-abliterated is a refusal-ablated open-weight language model, published on Hugging Face by mradermacher. It is 24B parameters and 24GB VRAM class class. It is distributed across 4 repositories in GGUF and GGUF (imatrix) formats, totalling 7.8K downloads.

Abliteration (directional ablation)24BMistral10B – 25B↓ 7,810♥ 5
View on Hugging Face ↗

Specification

Model name
Dolphin3.0-R1-Mistral-24B-abliterated
Source repository
mradermacher/Dolphin3.0-R1-Mistral-24B-abliterated-GGUF
Base model
Not declared
Publisher
mradermacher
Parameters
24B
Hardware class
10B – 25B — 24GB VRAM class
Licence
Not declared
Task
Unspecified
Context window
Not documented
Ablation scope
Not stated by the publisher
Formats available
GGUF, GGUF (imatrix)
First indexed
10 Feb 2025
Last updated
4 Apr 2026

Ablation technique

How Dolphin3.0-R1-Mistral-24B-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

“weighted/imatrix quants are available at https://huggingface.co/mradermacher/Dolphin3.0-R1-Mistral-24B-abliterated-i1-GGUF”

Running it

Commands are templates — confirm the exact repository and quant file before use.

llama.cpp / GGUF
hf download mradermacher/Dolphin3.0-R1-Mistral-24B-abliterated-GGUF --local-dir ./dolphin3.0-r1-mistral-24b-abliterated
llama-cli -m ./dolphin3.0-r1-mistral-24b-abliterated/<file>.gguf -p "..."
Transformers
from transformers import AutoModelForCausalLM, AutoTokenizer

repo = "mradermacher/Dolphin3.0-R1-Mistral-24B-abliterated-GGUF"
tok = AutoTokenizer.from_pretrained(repo)
model = AutoModelForCausalLM.from_pretrained(
    repo, torch_dtype="auto", device_map="auto"
)
vLLM
vllm serve mradermacher/Dolphin3.0-R1-Mistral-24B-abliterated-GGUF --trust-remote-code

Downloads and variants (3)

Every published repository of Dolphin3.0-R1-Mistral-24B-abliterated, including quantised re-releases by other authors.

RepositoryFormatDownloads
mradermacher/Dolphin3.0-R1-Mistral-24B-abliterated-GGUFsourceGGUF7.2K
mradermacher/Dolphin3.0-R1-Mistral-24B-abliterated-i1-GGUFGGUF (imatrix)444
lactroiii/Dolphin3.0-R1-Mistral-24B-abliterated-GGUFGGUF154

Security-team assessment

Derived from this model’s metadata and the publisher’s claims — not from benchmarks run by this site.

  • Dolphin3.0-R1-Mistral-24B-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.
  • Capability is inherited from its base model, not added by ablation — at 24B 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.
  • No licence is declared on this repository, which is a compliance problem in itself. Resolve it against the base model’s terms before use.

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 Dolphin3.0-R1-Mistral-24B-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 mradermacher, and does not independently verify publisher claims. See responsible use.