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

Athene Phi 3.5 Mini Instruct Orpo by EpistemeAI

 »  All LLMs  »  EpistemeAI  »  Athene Phi 3.5 Mini Instruct Orpo   URL Share it on

  9   
  Arxiv:2404.14219   4bit   Autotrain compatible   En   Endpoints compatible   Instruct   Llama   Pytorch   Quantized   Region:us   Safetensors   Sharded   Trl   Unsloth
Model Card on HF 🤗: https://huggingface.co/EpistemeAI/Athene-Phi-3.5-mini-instruct-orpo 

Athene Phi 3.5 Mini Instruct Orpo Benchmarks

LLME Score: 0.20602
nn.n% — How the model compares to the reference models: Anthropic Sonnet 3.5 ("so35"), GPT-4o ("gpt4o") or GPT-4 ("gpt4").
Athene Phi 3.5 Mini Instruct Orpo (EpistemeAI/Athene-Phi-3.5-mini-instruct-orpo)

Athene Phi 3.5 Mini Instruct Orpo Parameters and Internals

Model Type 
text generation
Use Cases 
Areas:
Research, Commercial applications
Applications:
General purpose AI systems, Applications requiring memory/compute constrained environments, Latency bound scenarios, Strong reasoning applications (e.g., code, math, and logic)
Primary Use Cases:
Accelerating research on language and multimodal models, Building block for generative AI powered features
Limitations:
Not designed or evaluated for all downstream purposes, Performance disparities across languages, Potentially generate inaccurate information
Considerations:
Evaluate and mitigate for accuracy, safety, and fairness. Adhere to applicable laws or regulations.
Additional Notes 
This model was trained 2x faster using Unsloth and Huggingface's TRL library.
Supported Languages 
Arabic (Supported), Chinese (Supported), Czech (Supported), Danish (Supported), Dutch (Supported), English (Supported), Finnish (Supported), French (Supported), German (Supported), Hebrew (Supported), Hungarian (Supported), Italian (Supported), Japanese (Supported), Korean (Supported), Norwegian (Supported), Polish (Supported), Portuguese (Supported), Russian (Supported), Spanish (Supported), Swedish (Supported), Thai (Supported), Turkish (Supported), Ukrainian (Supported)
Training Details 
Data Sources:
Phi-3 synthetic data, filtered publicly available websites
Data Volume:
3.4 trillion tokens
Methodology:
supervised fine-tuning, proximal policy optimization, direct preference optimization
Context Length:
128000
Training Time:
10 days
Hardware Used:
512 H100-80G GPUs
Model Architecture:
dense decoder-only Transformer model
Safety Evaluation 
Methodologies:
Not specified
Findings:
Multilingual performance and safety gaps, Representation of Harms & Perpetuation of Stereotypes, Inappropriate or Offensive Content, Information Reliability, Limited Scope for Code, Long Conversation
Risk Categories:
Quality of Service, Multilingual performance and safety gaps, Representation of Harms & Perpetuation of Stereotypes, Inappropriate or Offensive Content, Information Reliability, Limited Scope for Code, Long Conversation
Ethical Considerations:
Developers should implement additional safeguards at the application level and deploy models with appropriate mitigation measures to address potential biases and risks.
Responsible Ai Considerations 
Fairness:
Models may over- or under-represent groups of people, erase representation of some groups, or reinforce negative stereotypes.
Transparency:
Developers should inform end-users they are interacting with an AI system and follow transparency best practices.
Accountability:
Developers are accountable for the model's outputs within their specific use case and cultural, linguistic context.
Mitigation Strategies:
Fine-tune models for the specific use case, leverage language-specific safeguards, and perform regular assessments of high-risk scenarios.
Input Output 
Input Format:
Chat format prompts
Accepted Modalities:
Text
Output Format:
Generated text in response to input
Performance Tips:
Use the chat format for best prompt outputs.
Release Notes 
Date:
August 2024
Notes:
Update over the June 2024 instruction-tuned Phi-3 Mini release, based on user feedback, focused on multilingual, multi-turn conversation quality, and reasoning capability improvements.
LLM NameAthene Phi 3.5 Mini Instruct Orpo
Repository 🤗https://huggingface.co/EpistemeAI/Athene-Phi-3.5-mini-instruct-orpo 
Base Model(s)  unsloth/phi-3.5-mini-instruct-bnb-4bit   unsloth/phi-3.5-mini-instruct-bnb-4bit
Model Size3.8b
Required VRAM7.6 GB
Updated2025-02-26
MaintainerEpistemeAI
Model Typellama
Instruction-BasedYes
Model Files  5.0 GB: 1-of-2   2.6 GB: 2-of-2
Supported Languagesen
Gated ModelYes
Quantization Type4bit
Model ArchitectureLlamaForCausalLM
Licenseproprietary
Context Length131072
Model Max Length131072
Transformers Version4.44.2
Tokenizer ClassLlamaTokenizer
Padding Token<|placeholder6|>
Vocabulary Size32064
Torch Data Typefloat16

Best Alternatives to Athene Phi 3.5 Mini Instruct Orpo

Best Alternatives
Context / RAM
Downloads
Likes
DeepPhi 3.5 Mini Instruct128K / 7.6 GB550
Model128K / 7.6 GB410
Phi 3 5 Mini 3k Each128K / 7.6 GB680
Tara 3.8B146K / 7.6 GB370
Phi 3.5 Mini Instruct128K / 7.6 GB1478044
Superthoughts Mini V1.3.8B128K / 7.6 GB781
Phi 3 5 Mini Kp 12k Cfr Sft128K / 7.6 GB570
Phi 3 5 Mini Tictactoe1200128K / 7.6 GB100
...3 Mini 128K Instruct LLaMAfied128K / 7.6 GB173
Llamaphi 3 128K Instruct128K / 7.6 GB101
Note: green Score (e.g. "73.2") means that the model is better than EpistemeAI/Athene-Phi-3.5-mini-instruct-orpo.

Rank the Athene Phi 3.5 Mini Instruct Orpo 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? 44887 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