CUSTOM
CLANKERS.
← Browse all models

failspy / Model details

Phi-3-medium-4k-instruct-abliterated-v3

failspy/Phi-3-medium-4k-instruct-abliterated-v3

abliterated
Download files on Hugging Face ↗

Read the original model card ↗ · Choosing a model

About this model

failspy's abliterated derivative of microsoft/Phi-3-medium-4k-instruct, provided as bfloat16 safetensors weights under an MIT license. The creator says the weights were orthogonalized using a refined methodology based on the 'Refusal in LLMs is mediated by a single direction' work, aiming to inhibit the model's ability to express refusal while otherwise behaving the same as the original instruct model. According to the model card this is not a fine-tune, and the change is not guaranteed to prevent refusals; the model may still refuse or lecture about ethics and safety. The creator reports that in their experience this v3 is more likely to hallucinate than the original but generally as stable, and that the refined methodology seems to induce fewer hallucinations than their earlier versions. These behavior claims are the creator's own and have not been independently tested here.

Description sources

Source-based description, not an independent evaluation. Check the linked sources for the creator’s claims and limitations.

Size and formats

Base model
Unknown
Parameters
13.9602B
Main repository formats
safetensors
License tag
mit
Languages
multilingual
Last updated
2024-05-23 · 03:12:59.000 UTC

What changed

Model types are creator-reported, not independently tested. These labels describe source metadata, not a guarantee of behavior.

  • abliteratedmodel name: Phi-3-medium-4k-instruct-abliterated-v3

Family evidence

These groupings are source-reported. Explicit base identifiers take priority over generic architecture tags.

  • Phi: Family / architecture tag: phi3

Available versions

Versions appear here only when they explicitly name this exact model as their source. Check each repository for available files.

Link to this listing

Made or quantized this model? Paste one of these into your model card or README so people can find this listing.

Markdown

[Listed on Custom Clankers](https://customclankers.com/models/failspy-phi-3-medium-4k-instruct-abliterated-v3-4655a61fc814/)

HTML

<a href="https://customclankers.com/models/failspy-phi-3-medium-4k-instruct-abliterated-v3-4655a61fc814/">Listed on Custom Clankers</a>

Plain text links only: no image, no script, nothing loaded from this site.

Customize or build on this model

Listed from public GitHub data, not tested by us. Read the code and license before you run anything.

Tools to customize it further

Remove more refusals

  • Heretic ↗ ★ 36,365

    Removes refusal behavior from transformer models automatically. It pairs directional ablation (abliteration) with an Optuna parameter search that tries to cut refusals while keeping the model close to the original, so you don't have to pick layers by hand.

  • OBLITERATUS ↗ ★ 8,941

    A larger refusal-removal toolkit with several abliteration methods, usable locally or through a hosted Hugging Face Space. Check the telemetry setting before you run it: the README describes telemetry as opt-in, but on by default in the Spaces version.

Fine-tune it on your own data

  • Unsloth ↗ ★ 78,505

    A desktop app and Python library for running and training models locally. Its README lists LoRA, QLoRA, full fine-tuning, DPO and GRPO, plus export to formats including GGUF, on Windows, Linux, WSL and macOS.

  • LlamaFactory ↗ ★ 75,708

    Fine-tunes more than 100 language and vision-language models from a command line or a browser-based Web UI, aimed at training without writing code.

Make a smaller download (quantize)

  • llama.cpp ↗ ★ 132,209

    Runs models locally and includes the tools behind most GGUF files. A conversion script turns Hugging Face weights into GGUF, and llama-quantize shrinks a high-precision GGUF to lower-bit formats, which can cost some accuracy.

  • MLX LM ↗ ★ 7,461

    Apple's MLX package for running and fine-tuning models on Apple silicon. It can quantize models and upload them to the Hugging Face Hub, and supports LoRA and full fine-tuning.

Merge it with another model

  • mergekit ↗ ★ 7,414

    Merges pretrained models with many merge methods. It works out-of-core, so merges can run on a CPU or with as little as 8 GB of VRAM, and it also handles LoRA extraction and mixture-of-experts merges.

Learn the techniques

  • LLM Course ↗ ★ 84,653

    A free course with roadmaps and Colab notebooks. Its notebooks include one-click tools for merging with MergeKit, quantizing to GGUF, GPTQ, EXL2, AWQ and HQQ, and automatic abliteration with custom datasets.

Compare all tools →

Source provenance

Downloads
102
Likes
24
Metadata checked
2026-10-24 · 16:13:40 UTC
First seen in index
2026-10-24 · 16:13:40 UTC
Source revision
959b09eacf6cae85a8eb21b25e998addc89a367b

Source-card excerpt

Source model card excerpt: Phi-3-medium-4k-instruct-abliterated-v3 My Jupyter "cookbook" to replicate the methodology can be found here, refined library coming soon Phi-3-abliterated statement Took me a while to wizard

From failspy/Phi-3-medium-4k-instruct-abliterated-v3. Creator model-card excerpt, not an independent review.

Inspect before you run

Low refusal is not proof of correctness. Outputs can be inaccurate, harmful, or biased. Review the model card, license and files yourself. A license tag is not commercial-use advice. Parameter counts are source metadata; no hardware compatibility is implied.

This directory does not host or redistribute model weights. Refusal profiles are measured by us on a fixed prompt set; all other behavior claims are source-reported.

Collection rules & limitations ↗