LLM EXPLORER
 LLM Explorer: A Curated Large Language Model Directory and Analytics
Dark Theme
  LLM  List    LLM Hosting    LLM Leaderboards    Blog    Newsfeed    Advertise

Phi 3 Mini 128K Instruct Abliterated V3 by failspy

 »  All LLMs  »  failspy  »  Phi 3 Mini 128K Instruct Abliterated V3   URL Share it on

  1367   
  Autotrain compatible   Code   Conversational   Custom code   Endpoints compatible   Instruct   Multilingual   Phi3   Region:us   Safetensors   Sharded   Tensorflow
Model Card on HF 🤗: https://huggingface.co/failspy/Phi-3-mini-128k-instruct-abliterated-v3 

Phi 3 Mini 128K Instruct Abliterated V3 Benchmarks

LLME Score: 0.19467
nn.n% — How the model compares to the reference models: Anthropic Sonnet 3.5 ("so35"), GPT-4o ("gpt4o") or GPT-4 ("gpt4").
Phi 3 Mini 128K Instruct Abliterated V3 (failspy/Phi-3-mini-128k-instruct-abliterated-v3)
🌟 Advertise your project 🚀

Phi 3 Mini 128K Instruct Abliterated V3 Parameters and Internals

Model Type 
text-generation
Use Cases 
Primary Use Cases:
Research, Understanding orthogonalization effects
Limitations:
Model may refuse certain tasks due to inherent limitations, despite orthogonalization.
Additional Notes 
The model has had certain weights manipulated to inhibit refusal capabilities while maintaining original model behavior. Known to have quirks due to novel methodology.
Supported Languages 
multilingual ()
Training Details 
Methodology:
Orthogonalized bfloat16 safetensor weights. Ablation method based on single direction control of refusal features.
Model Architecture:
Inherits from original 70B instruct model with refusal directions orthogonalized out.
LLM NamePhi 3 Mini 128K Instruct Abliterated V3
Repository 🤗https://huggingface.co/failspy/Phi-3-mini-128k-instruct-abliterated-v3 
Model Size3.8b
Required VRAM7.7 GB
Updated2025-06-16
Maintainerfailspy
Model Typephi3
Instruction-BasedYes
Model Files  5.0 GB: 1-of-2   2.7 GB: 2-of-2
Model ArchitecturePhi3ForCausalLM
Licensemit
Context Length131072
Model Max Length131072
Transformers Version4.42.0.dev0
Tokenizer ClassLlamaTokenizer
Padding Token<|endoftext|>
Vocabulary Size32064
Torch Data Typebfloat16

Best Alternatives to Phi 3 Mini 128K Instruct Abliterated V3

Best Alternatives
Context / RAM
Downloads
Likes
Phi 4 Mini Instruct128K / 7.7 GB235092505
Phi 3.5 Mini Instruct128K / 7.7 GB276132876
Phi 3 Mini 128K Instruct128K / 7.7 GB3322061648
NuExtract V1.5128K / 7.7 GB10851189
NuExtract 1.5128K / 7.7 GB24309232
Phi 4 Mini Instruct128K / 7.7 GB451416
Phi 3.5 Mini TitanFusion 0.1128K / 7.7 GB250
Phi 4 Bifrost Sol 3.8B128K / 7.7 GB611
ECE EIFFEL 3Bv2128K / 7.7 GB50
Glider128K / 15.4 GB139041
Note: green Score (e.g. "73.2") means that the model is better than failspy/Phi-3-mini-128k-instruct-abliterated-v3.

Rank the Phi 3 Mini 128K Instruct Abliterated V3 Capabilities

🆘 Have you tried this model? Rate its performance. This feedback would greatly assist ML community in identifying the most suitable model for their needs. Your contribution really does make a difference! 🌟

Instruction Following and Task Automation  
Factuality and Completeness of Knowledge  
Censorship and Alignment  
Data Analysis and Insight Generation  
Text Generation  
Text Summarization and Feature Extraction  
Code Generation  
Multi-Language Support and Translation  

What open-source LLMs or SLMs are you in search of? 48185 in total.

Email us: info@extractum.io. Our Privacy Policy  | Terms and Conditions  | Suggest an improvement.
Our Social Media →  
Original data from HuggingFace, OpenCompass and various public git repos.
Release v20241124
  Support LLM Explorer