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

Meta Llama 3 8B Instruct Abliterated V3 by failspy

 »  All LLMs  »  failspy  »  Meta Llama 3 8B Instruct Abliterated V3   URL Share it on

  7320   
  Autotrain compatible   Conversational   Endpoints compatible   Instruct   Llama   Region:us   Safetensors   Sharded   Tensorflow
Model Card on HF 🤗: https://huggingface.co/failspy/Meta-Llama-3-8B-Instruct-abliterated-v3 

Meta Llama 3 8B Instruct Abliterated V3 Benchmarks

MMLU Pro: 29.48
GPQA: 1.9
MUSR: 2.84
BBH: 27.34
IFEval: 72.45 vs 88 (so35)-17.7%
ARC: 61.6 vs 96.7 (so35)-36.3%
HellaSwag: 78.35 vs 95.3 (gpt4)-17.8%
MMLU: 66.67 vs 88.3 (so35)-24.5%
TruthfulQA: 52.44 vs 59 (gpt4)-11.1%
WinoGrande: 74.51 vs 87.5 (gpt4)-14.8%
GSM8K: 70.05 vs 96.4 (so35)-27.3%
MATH Lvl 5: 9.59
LLME Score: 0.30066
nn.n% — How the model compares to the reference models: Anthropic Sonnet 3.5 ("so35"), GPT-4o ("gpt4o") or GPT-4 ("gpt4").
Meta Llama 3 8B Instruct Abliterated V3 (failspy/Meta-Llama-3-8B-Instruct-abliterated-v3)

Meta Llama 3 8B Instruct Abliterated V3 Parameters and Internals

Model Type 
text generation
Additional Notes 
The model has undergone orthogonalization to remove refusal behavior without altering other aspects of the original 70B instruct model. Known as "abliteration," this process differentiates it from standard uncensored fine-tunes. Potential quirks due to new methodology are acknowledged.
Training Details 
Methodology:
Orthogonalization (Ablation) of refusal behavior based on a refined methodology.
LLM NameMeta Llama 3 8B Instruct Abliterated V3
Repository 🤗https://huggingface.co/failspy/Meta-Llama-3-8B-Instruct-abliterated-v3 
Model Size8b
Required VRAM16.1 GB
Updated2025-05-31
Maintainerfailspy
Model Typellama
Instruction-BasedYes
Model Files  5.0 GB: 1-of-4   5.0 GB: 2-of-4   4.9 GB: 3-of-4   1.2 GB: 4-of-4
Model ArchitectureLlamaForCausalLM
Licensellama3
Context Length8192
Model Max Length8192
Transformers Version4.40.1
Tokenizer ClassPreTrainedTokenizerFast
Vocabulary Size128256
Torch Data Typebfloat16

Quantized Models of the Meta Llama 3 8B Instruct Abliterated V3

Model
Likes
Downloads
VRAM
...truct Abliterated V3 GPTQ 4bit0445 GB

Best Alternatives to Meta Llama 3 8B Instruct Abliterated V3

Best Alternatives
Context / RAM
Downloads
Likes
...otron 8B UltraLong 4M Instruct4192K / 32.1 GB3498107
UltraLong Thinking4192K / 16.1 GB3002
...a 3.1 8B UltraLong 4M Instruct4192K / 32.1 GB17624
...a 3.1 8B UltraLong 2M Instruct2096K / 32.1 GB8759
...otron 8B UltraLong 2M Instruct2096K / 32.1 GB39015
Zero Llama 3.1 8B Beta61048K / 16.1 GB7071
...otron 8B UltraLong 1M Instruct1048K / 32.1 GB180343
...a 3.1 8B UltraLong 1M Instruct1048K / 32.1 GB138729
...dger Nu Llama 3.1 8B UltraLong1048K / 16.2 GB573
....1 1million Ctx Dark Planet 8B1048K / 32.3 GB942
Note: green Score (e.g. "73.2") means that the model is better than failspy/Meta-Llama-3-8B-Instruct-abliterated-v3.

Rank the Meta Llama 3 8B 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? 47753 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 v20241227
  Support LLM Explorer