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

GPT Neox 20B by EleutherAI

 »  All LLMs  »  EleutherAI  »  GPT Neox 20B   URL Share it on

  28523   
  Arxiv:2101.00027   Arxiv:2104.09864   Arxiv:2201.07311   Arxiv:2204.06745   Autotrain compatible   Dataset:eleutherai/pile   En   Endpoints compatible   Gpt neox   Pytorch   Region:us   Safetensors   Sharded   Tensorflow
Model Card on HF 🤗: https://huggingface.co/EleutherAI/gpt-neox-20b 

GPT Neox 20B Benchmarks

MMLU Pro: 1.73
MUSR: 2.82
BBH: 4.93
IFEval: 25.87 vs 88 (so35)-70.6%
ARC: 45.73 vs 96.7 (so35)-52.7%
HellaSwag: 73.45 vs 95.3 (gpt4)-22.9%
MMLU: 25 vs 88.3 (so35)-71.7%
TruthfulQA: 31.61 vs 59 (gpt4)-46.4%
WinoGrande: 68.9 vs 87.5 (gpt4)-21.3%
GSM8K: 5.46 vs 96.4 (so35)-94.3%
MATH Lvl 5: 0.68
LLME Score: 0.12653
nn.n% — How the model compares to the reference models: Anthropic Sonnet 3.5 ("so35"), GPT-4o ("gpt4o") or GPT-4 ("gpt4").
GPT Neox 20B (EleutherAI/gpt-neox-20b)

GPT Neox 20B Parameters and Internals

Model Type 
Transformer-based, Autoregressive, Language Model
Use Cases 
Areas:
Research, Development
Primary Use Cases:
Feature extraction, downstream task learning
Limitations:
Not for direct deployment, Possible biases and inaccuracies
Additional Notes 
Not deduplicated before training, which might influence the model's output integrity.
Supported Languages 
English (fluent)
Training Details 
Data Sources:
The Pile
Data Volume:
825 GiB
Methodology:
Auto-regressive training using the GPT-NeoX library
Context Length:
2048
Training Time:
Approx. 150,000 steps with 1538 sequences per step
Model Architecture:
Similar to GPT-3, with specifics available in the paper.
Safety Evaluation 
Risk Categories:
misinformation, bias
Ethical Considerations:
Various biases related to gender, religion, and race as discussed in the Pile paper.
Responsible Ai Considerations 
Fairness:
See Pile paper for discussion on biases
Input Output 
Input Format:
Text input required for prompts
Accepted Modalities:
text
Output Format:
Autoregressive next-token predictions
LLM NameGPT Neox 20B
Repository 🤗https://huggingface.co/EleutherAI/gpt-neox-20b 
Model Size20b
Required VRAM40.8 GB
Updated2025-02-05
MaintainerEleutherAI
Model Typegpt_neox
Model Files  0.9 GB: 1-of-46   0.9 GB: 2-of-46   0.9 GB: 3-of-46   0.9 GB: 4-of-46   0.9 GB: 5-of-46   0.9 GB: 6-of-46   0.9 GB: 7-of-46   0.9 GB: 8-of-46   0.9 GB: 9-of-46   0.9 GB: 10-of-46   0.9 GB: 11-of-46   0.9 GB: 12-of-46   0.9 GB: 13-of-46   0.9 GB: 14-of-46   0.9 GB: 15-of-46   0.9 GB: 16-of-46   0.9 GB: 17-of-46   0.9 GB: 18-of-46   0.9 GB: 19-of-46   0.9 GB: 20-of-46   0.9 GB: 21-of-46   0.9 GB: 22-of-46   0.9 GB: 23-of-46   0.9 GB: 24-of-46   0.9 GB: 25-of-46   0.9 GB: 26-of-46   0.9 GB: 27-of-46   0.9 GB: 28-of-46   0.9 GB: 29-of-46   0.9 GB: 30-of-46   0.9 GB: 31-of-46   0.9 GB: 32-of-46   0.9 GB: 33-of-46   0.9 GB: 34-of-46   0.9 GB: 35-of-46   0.9 GB: 36-of-46   0.9 GB: 37-of-46   0.9 GB: 38-of-46   0.9 GB: 39-of-46   0.9 GB: 40-of-46   0.9 GB: 41-of-46   0.9 GB: 42-of-46   0.9 GB: 43-of-46   0.9 GB: 44-of-46   0.6 GB: 45-of-46   0.6 GB: 46-of-46
Supported Languagesen
Model ArchitectureGPTNeoXForCausalLM
Licenseapache-2.0
Context Length2048
Model Max Length2048
Transformers Version4.19.0.dev0
Vocabulary Size50432
Torch Data Typefloat16

Best Alternatives to GPT Neox 20B

Best Alternatives
Context / RAM
Downloads
Likes
GPT NeoXT Chat Base 20B2K / 41.2 GB4048696
EleutherAI GPT Neox 20B 4bits2K / 12.5 GB50
...t Gm Oasst1 Multilang 1024 20B2K / 41.2 GB136910
H2ogpt Gm Oasst1 En 1024 20B2K / 41.2 GB13334
H2ogpt Oasst1 512 20B2K / 41.2 GB135840
GPT Neox 20B Full Precision2K / 82.5 GB13420
Oasst GPT Neox 20B 1000 Steps2K / 41.2 GB13240
Oasst GPT Neox 20B 3000 Steps2K / 41.2 GB13290
GPTNeoX 20B TestGen Dart V1.02K / 41.2 GB142
GPT NeoX 20B Erebus2K / 41.4 GB352184
Note: green Score (e.g. "73.2") means that the model is better than EleutherAI/gpt-neox-20b.

Rank the GPT Neox 20B 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? 42577 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