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

Zephyr 7B Alpha by HuggingFaceH4

 »  All LLMs  »  HuggingFaceH4  »  Zephyr 7B Alpha   URL Share it on

  13539   
  Arxiv:2305.14233   Arxiv:2305.18290   Arxiv:2310.01377   Arxiv:2310.16944   Autotrain compatible Base model:finetune:mistralai/... Base model:mistralai/mistral-7...   Conversational   Dataset:openbmb/ultrafeedback   Dataset:stingning/ultrachat   En   Endpoints compatible   Generated from trainer   Mistral   Pytorch   Region:us   Safetensors   Sharded   Tensorflow
Model Card on HF 🤗: https://huggingface.co/HuggingFaceH4/zephyr-7b-alpha 

Zephyr 7B Alpha Benchmarks

LMSys ELO: 1042 vs 1272 (so35)-18.1%
MMLU Pro: 19.94
GPQA: 6.38
MUSR: 7.5
BBH: 23.96
IFEval: 51.91 vs 88 (so35)-41%
ARC: 61.01 vs 96.7 (so35)-36.9%
HellaSwag: 84.04 vs 95.3 (gpt4)-11.8%
MMLU: 61.39 vs 88.3 (so35)-30.5%
TruthfulQA: 57.9 vs 59 (gpt4)-1.9%
WinoGrande: 78.61 vs 87.5 (gpt4)-10.2%
GSM8K: 14.03 vs 96.4 (so35)-85.4%
MATH Lvl 5: 1.74
LLME Score: 0.41509
nn.n% — How the model compares to the reference models: Anthropic Sonnet 3.5 ("so35"), GPT-4o ("gpt4o") or GPT-4 ("gpt4").
Zephyr 7B Alpha (HuggingFaceH4/zephyr-7b-alpha)

Zephyr 7B Alpha Parameters and Internals

Model Type 
GPT-like, Text Generation
Use Cases 
Areas:
Research, Commercial Applications
Primary Use Cases:
Chat Applications
Limitations:
Can produce problematic outputs when prompted
Considerations:
Model can be tested via demo link for chat capabilities.
Additional Notes 
It can generate problematic text if prompted to do so.
Supported Languages 
en (Primarily supported)
Training Details 
Data Sources:
stingning/ultrachat, openbmb/UltraFeedback
Methodology:
Direct Preference Optimization (DPO)
Model Architecture:
GPT-like
Input Output 
Input Format:
Uses tokenizer's chat template to format messages.
Accepted Modalities:
text
Output Format:
Text outputs in the style prompted (e.g., pirate style)
Performance Tips:
Ensure format is appropriate for intended style and content.
LLM NameZephyr 7B Alpha
Repository 🤗https://huggingface.co/HuggingFaceH4/zephyr-7b-alpha 
Base Model(s)  mistralai/Mistral-7B-v0.1   mistralai/Mistral-7B-v0.1
Model Size7b
Required VRAM14.4 GB
Updated2024-12-22
MaintainerHuggingFaceH4
Model Typemistral
Model Files  1.9 GB: 1-of-8   1.9 GB: 2-of-8   2.0 GB: 3-of-8   1.9 GB: 4-of-8   2.0 GB: 5-of-8   1.9 GB: 6-of-8   2.0 GB: 7-of-8   0.8 GB: 8-of-8   1.9 GB: 1-of-8   1.9 GB: 2-of-8   2.0 GB: 3-of-8   1.9 GB: 4-of-8   2.0 GB: 5-of-8   1.9 GB: 6-of-8   2.0 GB: 7-of-8   0.8 GB: 8-of-8   0.0 GB
Supported Languagesen
Model ArchitectureMistralForCausalLM
Licensemit
Context Length32768
Model Max Length32768
Transformers Version4.34.0
Tokenizer ClassLlamaTokenizer
Padding Token</s>
Vocabulary Size32000
Torch Data Typebfloat16

Quantized Models of the Zephyr 7B Alpha

Model
Likes
Downloads
VRAM
Zephyr 7B Alpha GGUF13812423 GB
Zephyr 7B Alpha GGUF01152 GB
Zephyr 7B Alpha GPTQ2611624 GB
Zephyr 7B Alpha GGUF02122 GB
Zephyr 7B Alpha AWQ16834 GB

Best Alternatives to Zephyr 7B Alpha

Best Alternatives
Context / RAM
Downloads
Likes
...Nemo Instruct 2407 Abliterated1000K / 24.5 GB26689
MegaBeam Mistral 7B 512K512K / 14.4 GB711346
SpydazWeb AI HumanAI RP512K / 14.4 GB261
SpydazWeb AI HumanAI 002512K / 14.4 GB191
...daz Web AI ChatML 512K Project512K / 14.5 GB120
MegaBeam Mistral 7B 300K282K / 14.4 GB328215
Hebrew Mistral 7B 200K256K / 30 GB319315
Astral 256K 7B V2250K / 14.4 GB190
Astral 256K 7B250K / 14.4 GB160
Boptruth Agatha 7B128K / 14.4 GB4750

Rank the Zephyr 7B Alpha 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? 40066 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 v20241217
  Support LLM Explorer